switchyard

for agents: this page is markdown at https://switchyard.so/docs/parity.md; every page is listed in llms.txt.

Board, API, CLI and MCP

Everything the board does can be done without it: through the REST API, through the yard CLI, and for a project's own work through the MCP server agents use. This page lists every action on the board with its API route, its yard command and its MCP tool, and says why where one has none. A few actions stay on the board on purpose; a table says which, and why. The last section lists every route with its yard command and MCP tool.

Project routes are under https://app.switchyard.so/v1/projects/<project> and take a project's admin key, a task key, an agent group key or a signed-in person. Organization and account routes are under https://app.switchyard.so/api and act for a person signed in (yard login https://app.switchyard.so --github). The MCP server answers inside one project, even for an agent signed in as its person over OAuth, so organization and account actions have no MCP tool: they take a person's sign-in, which yard holds and an agent's key does not.

Tasks

On the boardAPIyardMCP
See main, the tasks and the merge queueGET /statusyard statusyard_status
See the project itself: its name, main branch, remote and the GitHub repository it followsGET /v1/projects/<project>yard project showyard_whoami
See the active tasks, and the merged and abandoned ones a page at a timeGET /workspaces, GET /workspaces?status=finishedyard branch list [--finished], yard status --allyard_status, yard_finished
See the organization's spending limit: the share used, the forecast and who can raise it (raising it is a person's, on the billing page)GET /spending-limityard api GET /v1/projects/{project}/spending-limit reads it with the project's key; yard org limit reads the organization's own limit (GET /api/orgs/:org/billing/limit) with your sign-inyard_spending_limit
Start a taskPOST /workspacesyard branch newyard_workspace_create
Give a task its planPUT /workspaces/:workspace/planyard planyard_plan
Tag a taskPOST /workspaces/:workspace/tagsyard tagyard_tag
Assign a task to a personPOST /workspaces/:workspace/opsyard assignyard_assign
Abandon a task, saying why if you likePOST /workspaces/:workspace/ops with op abandon and args.reasonyard abandon [--reason]yard_workspace_op
Read what overlaps with a taskGET /workspaces/:workspace/warningsyard warningsyard_warnings
Read Switchyard's note on a task: what it last changed, its merge queue place, its conflictsGET /workspaces/:workspace/noticeyard noticeyard_notice
Submit a taskPOST /workspaces/:workspace/submityard submityard_submit
Merge a task first, as a hotfix (no board screen)POST /workspaces/:workspace/opsyard hotfixyard_priority
Merge a task in the priority lane, after hotfixes and before the normal lane (no board screen)POST /workspaces/:workspace/opsyard priorityyard_priority
Replace a task's keyPOST /workspaces/:workspace/keyyard branch keyyard_task_key

Review, conflicts and history

On the boardAPIyardMCP
Read the review inboxGET /statusyard reviewyard_review
Approve a change, ask for a fix or comment, bound to the commit you read (refused if the version now names another)POST /review/:change/:version with commityard review approve, yard review request-fix, yard review comment (send the inbox item's commit)yard_review with commit
Read a change, its versions and its diffGET /changes/:change, GET /changes/:change/interdiffyard showyard_change
See the conflicts and the steps triedGET /conflictsyard conflictsyard_conflicts
Read each side and the base of a conflicted fileGET /conflicts/:change/sidesyard conflicts --sidesyard_conflict_sides
Fix a conflictPOST /conflicts/:change/resolveyard resolveyard_resolve
Read the project's historyGET /oplogyard logyard_log
Undo an actionPOST /oplog/:seq/undoyard undoyard_undo
Find which change wrote a lineGET /whyyard whyyard_why
Check a merged commit's signed merge recordGET /attestations/:commityard verifyyard_attestation
Read the project's signing key, to pin it in .yard/allowed_signersGET /signeryard verify --signeryard_signer
Read a check's full logGET /checks/logs/<id>yard check logyard_check_log
Run a command in a failed check's kept machinePOST /checks/<id>/shellyard check shell <check> -cyard_check_shell
End a kept machineDELETE /checks/<id>/shellyard check shell <check> --doneyard_check_shell with done
Read the project's numbers over timeGET /insightsyard insightsyard_insights
See the merge queue's healthGET /queue-healthyard queue healthyard_queue_health
Tell the Switchyard team an answer confused an agent (no board screen; read on the admin page)POST /feedbackyard feedbackyard_feedback

Running the project

On the boardAPIyardMCP
See whether Switchyard, GitHub and Cloudflare are working (the live indicator)GET /v1/upstreamyard status (its first line, when something is wrong)yard_status (its upstream)
Read the settingsGET /settingsyard settingsyard_settings
Turn reviews on or offPUT /reviewsyard settings set reviewsyard_setting
Let conflict fixes merge by themselves, or wait for approvalPUT /fixesyard settings set fixesyard_setting
Choose who submits a taskPUT /task-permissionyard settings set who-submitsyard_setting
Choose whether opening the project wakes the Git enginePUT /wake-on-openyard settings set wake-on-openyard_setting
Choose what the checks may reach on the networkPUT /checks/networkyard settings set check-networkyard_setting
Choose whether a task's failed check keeps its machinePUT /keep-failed-machinesyard settings set keep-failed-machinesyard_setting
Choose whether the Git engine stays warm for an hour after workPUT /warm-houryard settings set warm-houryard_setting
Choose whether the person whose task it is may approve its protected changesPUT /self-approvalyard settings set self-approvalyard_setting
Choose the apps that write, such as renovate[bot] (adding one is a person's, in a browser)PUT /allowed-appsyard settings set allowed-appsyard_setting
Choose whether checks keep their package downloads between runsPUT /build-cacheyard settings set build-cacheyard_setting
Choose whether a task waiting for review runs its checks while it waitsPUT /checks-while-reviewingyard settings set checks-while-reviewingyard_setting
Choose whether a task's checks run ahead of submit once its pushes go quietPUT /checks-before-submityard settings set checks-before-submityard_setting
Choose whether the Git engine grows to a larger machine under load (hosted: the organization's choice, and letting it grow is a person's, in a browser)PUT /engine-sizingyard settings set engine-sizingyard_setting
Choose whether the integration is rebuilt only where tasks movedPUT /incremental-rebuildyard settings set incremental-rebuildyard_setting
Choose whether the engine merges path by pathPUT /grouped-mergeyard settings set grouped-mergeyard_setting
Choose whether a push to a merged task continues it as a new changePUT /continue-mergedyard settings set continue-mergedyard_setting
Choose whether tasks that touch nothing in common ride separate merge trains (no control on the settings page yet; off for every project)PUT /merge-trainsyard settings set merge-trains off sends; on prints the link: where people sign in, turning it on is a person's act in a browser signed in within 15 minutes, so yard prints the project's settings page's link and exits 8 (approval_required); there is no control on the settings page yet, and Switchyard turns merge trains on for a project on requestyard_setting: off is done; on is answered approval_required with the project's settings page
Keep the rules and checks in SwitchyardPUT /rulesyard rules show, yard rules keepyard_rules_keep
Say a path needs no check ("this path needs no check")POST /rules/uncheckedyard rules uncheckedyard_unchecked
Choose whether a change only to documentation needs a checkPUT /docs-excusedyard settings set docs-excusedyard_setting
Start holding untested changes before the guard's seven days of watching endPOST /untested-guard/enforceyard rules enforceyard_untested_guard_enforce
Add the protected areas a project set up before them misses (projects that follow GitHub)GET /api/projects/<project>/protected-areas, POST with { ids }yard rules protect [--add]not yet: an agent's key has no person on GitHub to answer for the change
Turn a kept check off or back onPUT /checks/offyard check off, yard check onyard_checks_off
Try the checks that are off on main againPOST /checks/try, GET /checks/triedyard check try, yard check triedyard_checks_try, yard_checks_tried
Pause merging, and resume itPUT /pause, DELETE /pauseyard pause, yard resumeyard_pause, yard_resume
Copy the GitHub repository again after a failed first copyPOST /copy-againyard copy-againyard_copy_again
Cancel the copy of the repository's history behind main, and copy it againDELETE /copy-history, POST /copy-historyyard copy-history --cancel, yard copy-historyyard_copy_history
Wake the project's Git engine ahead of work, and see what the project is doingPOST /engine/wake, activity on GET /statusyard engine wake [--wait], yard status [--watch]yard_wake_engine, activity in yard_status
Set up a project for a repositoryPOST /v1/projectsyard initsigned-in person: making a project acts for a person across projects, and the MCP server works inside one project
Set a GitHub repository up: the App's installations, the repositories, the checks and rules drafted from its CI, then commit them or open a pull requestGET /api/onboarding/github, GET /api/onboarding/repositories, GET /api/onboarding/analysis, POST /api/onboarding/setupyard setup github, yard setup repos, yard setup analyze [--plan], yard setup applysigned-in person: it answers to a person's own access on GitHub; it stays allowed with the person's yard sign-in, so an agent sets a repository up for its person, but no key or MCP tool does
Choose where a project's checks run: Switchyard's machines or the repository's own GitHub ActionsPOST /api/onboarding/checks-onyard setup checks-onsigned-in person: the change answers to a person's own access on GitHub
Show the command that clones mainPOST /main/credentialsyard fetchyard_git_credentials with main
Rotate, disable, enable or revoke an agent group keyPOST /fleets/:fleet/:actionyard group <F> disable and yard group <F> revoke send, so a leaked key is stopped fast; yard group <F> rotate and yard group <F> enable print the link, since a person does them in a browser signed in within 15 minutes (exit 8, approval_required); of keys, only the project admin key itself still rotates and enablesyard_agent_group_key: revoke and disable are done; rotate and enable are answered approval_required with the project's settings page for an agent signed in over MCP and for every organization key
Make an agent group key, and list themPOST /fleets, GET /fleetsyard group list; yard group create prints the link, since a person makes a key in a browser signed in within 15 minutes (exit 8, approval_required); of keys, only the project admin key itself still makes oneyard_agent_groups; yard_agent_group_create is answered approval_required with the project's settings page for an agent signed in over MCP and for every organization key
Rotate the project's admin keyPOST /admin-key/rotateyard project rotate-admin-key --confirm prints the link: a person with admin access rotates it in a browser signed in within 15 minutes, so yard prints the project's settings page's link and exits 8 (approval_required)signed-in person: every earlier admin key stops at once, so Switchyard takes it only from a person with admin access to the project, in a browser signed in within 15 minutes, never from a key, the admin key itself included
Add, list and remove webhooksPOST /webhooks, GET /webhooks, DELETE /webhooks/:idyard webhook add, yard webhook list, yard webhook removeyard_webhooks
Send a webhook endpoint a test event and see its resultPOST /webhooks/:id/testyard webhook testyard_webhooks with action: "test"
See and forget remembered conflict resolutionsGET /resolutions, DELETE /resolutions/:keyyard resolutions, yard resolutions forgetyard_resolutions
Set up a coding agent on a computerPOST /api/setup-codesyard agent setupinternal: it is how an agent gets the MCP server, so the server has no tool for it
Schedule a project's deletion, or keep the projectPOST /api/projects/:project/deletion, DELETE /api/projects/:project/deletionyard project deletion reads it and yard project keep keeps the project; yard project delete prints the link: a person schedules a deletion in a browser signed in within 15 minutes, so yard prints the project's settings page's link and exits 8 (approval_required)signed-in person: an owner or admin of the organization schedules it in a browser, and keeps the project with their sign-in

Working in a checkout

These have no button on the board: they are how an agent or a person works in a task's own Git checkout.

WhatAPIyardMCP
Open a task in a checkout, with its Git keyPOST /workspaces/:workspace/credentialsyard branch open, yard credentialyard_git_credentials
Take a task as one of many agents sharing a keyPOST /workspaces/claimyard branch new with the agent group keyyard_workspace_claim
Say which paths and symbols a task will touchPOST /workspaces/:workspace/claimsyard claimyard_claim
Bring main, or the task below, into a taskPOST /workspaces/:workspace/opsyard syncyard_sync
Commit, and shape the task's stackGit push to the task, POST /workspaces/:workspace/opsyard commit, yard absorb, yard split, yard squash, yard move, yard uncommit, yard edityard_absorb, yard_split, yard_squash, yard_workspace_op
Send the pushes queued while Switchyard could not be reachednone: the queue lives in the checkout's .git and is sent with Gityard push --queued, and yard status --local lists itnone: an MCP agent pushes with plain Git, which queues nothing; its commits stay on its branch to push again once Switchyard answers
Run a coding agent in a tasknone: the agent runs on your own computeryard agent runinternal: it starts the agent that then uses the MCP server
Check a webhook request's signaturenone: the check runs on your own computer and sends nothing to Switchyardyard webhook verifynone: the signing secret stays on your computer, so the MCP server has no tool for it
Sign in, check the setup and the shellGET /api/meyard login, yard logout, yard whoami, yard doctor, yard completionsyard_whoami reads the key's or agent's own identity in the project (GET /me); GET /api/me is a signed-in person's, with no MCP tool

Organizations

On the boardAPIyard
List your organizationsGET /api/orgsyard org list
See the members and their teamsGET /api/orgs/:org/membersyard org members
List the projects you can seeGET /v1/projectsyard project list
Invite someone by GitHub loginPOST /api/invitationsyard org invite prints the link: inviting is a person's act in a browser signed in within 15 minutes, so yard prints the teams page's link and exits 8 (approval_required)
See and withdraw open invitationsGET /api/orgs/:org/invitations, POST /api/auth/organization/cancel-invitationyard org invitations, yard org withdraw
Answer an invitation sent to youGET /api/invitations, POST /api/auth/organization/accept-invitation, POST /api/auth/organization/reject-invitationyard org invited, yard org accept, yard org decline
Change a member's role, or remove them, keeping chosen keys for the teamGET /api/orgs/:org/members/:member/keys, POST /api/orgs/:org/members/:member/role, POST /api/orgs/:org/members/:member/removeyard org role --keep, yard org remove --keep print the link: a person does both in a browser signed in within 15 minutes, so yard prints the teams page's link and exits 8 (approval_required)
Leave an organizationPOST /api/orgs/:org/leaveyard org leave prints the link: a person leaves in a browser signed in within 15 minutes, so yard prints the teams page's link and exits 8 (approval_required)
Make, rotate, disable, enable, revoke and keep for the team API keysGET /api/orgs/:org/keys, POST /api/orgs/:org/keys, POST /api/orgs/:org/keys/:key/:actionyard org keys; yard org key disable and yard org key revoke send, so a leaked key is stopped fast; yard org key create, yard org key rotate, yard org key enable and yard org key keep print the link, since a person does them in a browser signed in within 15 minutes (exit 8, approval_required)
Turn AI conflict fixes on or offGET /api/orgs/:org/ai-fixes, PUT /api/orgs/:org/ai-fixesyard org ai-fixes reads them; yard org ai-fixes on prints the page, since turning them on or off is browser only: they spend on the bill, so a person does it in a browser signed in within 15 minutes
Read the month's usageGET /api/orgs/:org/usageyard org usage
Read a month's bill and every check run in itGET /api/orgs/:org/billing, GET /api/orgs/:org/billing/runs.csvyard org billing
Choose who else gets billing emails, beside the owners, admins and billing membersGET /api/orgs/:org/billing (its contacts), PUT /api/orgs/:org/billing/contactsyard org billing contacts reads them; with --add or --remove it prints the link: a person changes who gets the billing emails in a browser signed in within 15 minutes, so yard prints the billing page's link and exits 8 (approval_required)
Set a spending limit: an email at a warning amount; at the limit, pushes, new tasks and submits are kept but nothing is tested or merged, and after 24 hours or $2 of engine time, whichever comes first, new work is refused (clones and fetches keep working)GET /api/orgs/:org/billing/limit, PUT /api/orgs/:org/billing/limityard org limit; yard org limit set prints the billing page's link to confirm it in a browser
Raise the spending limit in one click: +$100, +$500, remove it, or allow until tomorrow (a person who looks after billing, in a browser, never a key or a yard sign-in)POST /api/orgs/:org/billing/limit/raise, GET /api/orgs/:org/billing/limit/offer for an email's linkbrowser only: a person who looks after billing, in a browser session from the last 15 minutes, never a key or a yard sign-in
Make and remove teams, and add and take out their members (owners and admins)POST /api/orgs/:org/teams, DELETE /api/orgs/:org/teams/:team, POST and DELETE /api/orgs/:org/teams/:team/membersyard org team create, yard org team delete, yard org team add and yard org team remove print the teams page's link: who is on a team decides who reviews what, so a person makes or removes a team, and adds or takes out a member, in a browser signed in within 15 minutes (exit 8, approval_required)
Rename an organization (owners only)POST /api/auth/organization/updateyard org rename
Delete an organization (owners only, once it owns no project and its billing has not started)POST /api/auth/organization/deleteyard org delete prints the page, since deleting is browser only: it cannot be undone, so its owner does it in a browser signed in within 15 minutes
Set or remove an organization secret for hosted agent runs (owners only; hosted agent runs are off today, and AI suggestions never read one)PUT /api/orgs/:org/secrets/:name, DELETE /api/orgs/:org/secrets/:nameyard org secret prints the page, since a secret is browser only: its owner types it in a browser signed in within 15 minutes, and no command line or chat ever carries it
See an organization, its plan and teamsGET /api/orgs/:org, GET /api/orgs/:org/teams, GET /api/orgs/:org/teams/:team/membersyard org show, yard org teams, yard org members --team
Read the organization's audit log by kind, person, outcome and dates, and export it (the audit log page)GET /api/orgs/:org/audit, with format=csv or format=jsonlyard org audit
See the projects the organization deleted lately, who asked, and until when their data is keptGET /api/orgs/:org/deleted-projectsyard org deleted
Make an organization (the sign-up step); an agent's account is refused, and on open sign-up so is an account whose email is not provenPOST /api/auth/organization/createyard org create prints the teams page's link: a new organization is a person's choice, so a person makes it in a browser where they are signed in, however long ago (exit 8, approval_required), and yard refuses an agent's or a script's run before sending; then yard init --organization makes a project in it
See the card and where payments stand, and add or change the card on Stripe's page (the first card pays the plan's minimum for the rest of the month)GET /api/orgs/:org/billing/card, POST /api/orgs/:org/billing/card with { plan? }yard org card reads it; yard org card add prints the link: a person adds or changes the card in a browser where they are signed in, however long ago, so yard prints the billing page's link and exits 8 (approval_required), and the person chooses there the plan whose minimum the first card pays
Read the statements and their invoices: what was charged, when, and what waits for review (no board screen)GET /api/orgs/:org/billing/invoicesyard org invoices
Choose whether the organization's Git engines grow to a larger machine under load, billed at that size, or stay small (owners and admins)GET /api/orgs/:org/engine-sizing, PUT /api/orgs/:org/engine-sizingyard org engine-sizing reads it and keeps them small; yard org engine-sizing your-org grows prints the page, since letting them grow is a person's, in a browser
See single sign-on: its connections, whether it is required and the DNS records still to addGET /api/orgs/:org/ssoyard org sso
Check a single sign-on connection's DNS records now (owners)POST /api/orgs/:org/sso/:provider/verifyyard org sso verify
Set up, remove or require single sign-on (owners; Scale and Enterprise)POST /api/orgs/:org/sso, DELETE /api/orgs/:org/sso/:provider, PUT /api/orgs/:org/sso-requiredbrowser only: an owner does it in a browser signed in within 15 minutes; yard org sso prints the page
See the SCIM directory: whether one is paired, with which connection, and how many people it holdsGET /api/orgs/:org/scimyard org scim
Pair, make the directory key again for, or unpair the SCIM directory (owners; Enterprise)POST /api/orgs/:org/scim, DELETE /api/orgs/:org/scimbrowser only: the directory key is a secret shown once, so an owner does it in a browser signed in within 15 minutes; yard org scim prints the page
See the two-factor level and how members are coveredGET /api/orgs/:org/two-factoryard org two-factor
Set the two-factor level: recommended or required (owners)PUT /api/orgs/:org/two-factor-requiredbrowser only: a security setting an owner changes in a browser signed in within 15 minutes; yard org two-factor prints the page
See the IP allowlist, and whether your address is in it (no board screen)GET /api/orgs/:org/ip-allowlistnone: a member reads it with their sign-in; yard api GET /api/orgs/your-org/ip-allowlist sends it
Set the IP allowlist (owners; no board screen)PUT /api/orgs/:org/ip-allowlistbrowser only: a security setting an owner changes in a browser signed in within 15 minutes, from an address the new list allows

Your account

On the boardAPIyard
See who you are signed in asGET /api/meyard whoami
Sign in, and sign out/api/auth/device/code, POST /api/auth/sign-outyard login --github, yard logout
See where you are signed in, and sign one or all outGET /api/sessions, DELETE /api/sessions/:id, POST /api/auth/revoke-sessionsyard account sessions, yard account sessions end <id>; yard account sessions end --all prints the link: signing out everywhere is a person's act in a browser where they are signed in, however long ago, so yard prints the account page's link and exits 8 (approval_required)
Choose which emails Switchyard sends youGET /api/me/emails, PUT /api/me/emailsyard account emails, yard account emails set
Set up two-factor sign-in: an authenticator app, its backup codes and passkeys/api/me/two-factor, /api/me/passkeysbrowser only: a person does it in a browser signed in within 15 minutes; no key and no agent does it
Report a bug from the board's side panelPOST /api/me/bug-reportsnone: it is a person's own report from a page; an agent tells the Switchyard team with yard feedback and yard_feedback
See and end the agents you signed in over MCPGET /api/me/agents, DELETE /api/me/agents/:agentyard account agents, yard account agents end
See your notifications, mark them read or clear themGET /api/me/notifications, POST /api/me/notifications/read, POST /api/me/notifications/clearyard notifications, yard notifications read, yard notifications clear; the bell marks what it showed seen, which yard never does. An agent signed in over MCP as its person reads that person's open reviews and tasks of one project with yard_notifications
Mute a task or a project, and lift itPOST /api/me/notifications/muteyard notifications mute, yard notifications unmute
Choose what reaches you and how, and the daily summaryGET /api/me/notification-settings, PUT /api/me/notification-settingsyard notifications settings, yard notifications settings set, yard notifications settings unset, yard notifications settings summary

Any route

ActionAPIyard
Send any route on this page with the key yard would use, for what no command does: a project's routes with the project's key, /api routes with your sign-inany routeyard api <METHOD> <path>, with {project} and {task} filled in; a change needs --yes off a terminal

On the board only, on purpose

ActionWhy it stays on the board
Let a project's conflict records help improve conflict fixingIt is consent, so a person gives it on the settings page, in a browser. No key, no yard sign-in and no agent can give it.
Delete your accountIt cannot be undone, so it takes a browser signed in within 15 minutes, never a yard sign-in.
Switchyard's own admin pagesThey are for the people who run Switchyard, not for organizations.

The website's waitlist and contact forms (/api/waitlist and /api/contact on https://switchyard.so) are a person's, in a browser, behind a bot check. They have no API route in the map, no yard command and no MCP tool.

Every MCP tool is on this page, and so is every yard command in the CLI page's command table. A test fails when one of them is added without its row, and when a row leaves a way out without saying why. The CLI page lists every command at every level but the ones only Switchyard's own team runs, so a third-level command such as yard org key create is checked too.

<!-- every route: rendered from the API map by node scripts/docs-reference.mjs -->

Every route

This table is rendered from the API map, the product's own list of its routes, which a test holds to the routes the code answers, so it changes when a route does. Paths are under https://app.switchyard.so. A route with no yard command or MCP tool says why: browser only is a person's own act in a browser signed in within 15 minutes (consent, the spending limit, billing contacts, approvals a person confirms, approving yard's sign-in, AI fixes, the organization's agent switches, deleting the account or an organization, organization secrets, single sign-on, the SCIM directory, security settings and a team's members), or, for making an organization, in a browser signed in at any time, that no key and no yard sign-in may do; signed-in person takes a person's sign-in, which yard holds after yard login --github, never a key, and which the MCP server, working inside one project with its keys, does not hold (people and roles, organizations, rotating an admin key); internal is the plumbing another way uses; no command means yard has no command of its own that calls the route, and yard api sends it with yard's key or sign-in. The routes Switchyard's own team uses and the steps of signing in are left out. The same routes are described for machines in the OpenAPI document.

Tasks

RouteyardMCP
GET /v1/projects/{project}/workspacesyard branch listyard_finished
POST /v1/projects/{project}/workspacesyard branch newyard_workspace_create
DELETE /v1/projects/{project}/workspacesinternal: a client that gives up a create it keeps sending (yard, the MCP server) calls it with the create's key itself; no person or agent asks for itinternal: a client that gives up a create it keeps sending (yard, the MCP server) calls it with the create's key itself; no person or agent asks for it
POST /v1/projects/{project}/workspaces/claiminternal: it serves the MCP server's yard_workspace_claim only; yard branch new starts a task by the route that creates oneyard_workspace_claim
DELETE /v1/projects/{project}/workspaces/claiminternal: a client that keeps sending a claim gives it up here with the claim's key; yard sends no claimsinternal: a client that keeps sending a claim gives it up here with the claim's key; yard_workspace_claim's withdraw gives up its own claims by the route that gives up a create instead
PUT /v1/projects/{project}/workspaces/{workspace}/planyard planyard_plan
POST /v1/projects/{project}/workspaces/{workspace}/tagsyard tagyard_tag
GET /v1/projects/{project}/workspaces/{workspace}/warningsyard warningsyard_warnings
GET /v1/projects/{project}/pulls/{number}yard pr statusyard_pr_status
GET /v1/projects/{project}/workspaces/{workspace}/noticeyard noticeyard_notice
POST /v1/projects/{project}/workspaces/{workspace}/observeyard commityard_warnings
POST /v1/projects/{project}/workspaces/{workspace}/claimsyard claimyard_claim
POST /v1/projects/{project}/workspaces/{workspace}/opsyard syncyard_workspace_op
POST /v1/projects/{project}/workspaces/{workspace}/credentialsyard branch openyard_git_credentials
POST /v1/projects/{project}/main/credentialsyard fetchyard_git_credentials
POST /v1/projects/{project}/workspaces/{workspace}/retrysigned-in person: a person decides that a check's failures with no cause found were not the task's, never an agent, so no MCP tool retries it; yard has no command for it: a person signed in with yard login calls the route with yard apisigned-in person: a person decides that a check's failures with no cause found were not the task's, never an agent, so no MCP tool retries it; yard has no command for it: a person signed in with yard login calls the route with yard api
POST /v1/projects/{project}/workspaces/{workspace}/submityard submityard_submit

Review, conflicts and resolutions

RouteyardMCP
GET /v1/projects/{project}/resolutionsyard resolutionsyard_resolutions
DELETE /v1/projects/{project}/resolutions/{key}yard resolutions forgetyard_resolutions
GET /v1/projects/{project}/changes/{change}yard showyard_change
GET /v1/projects/{project}/changes/{change}/interdiffyard showyard_change
GET /v1/projects/{project}/conflictsyard conflictsyard_conflicts
GET /v1/projects/{project}/conflicts/{change}/sidesyard conflicts --sidesyard_conflict_sides
POST /v1/projects/{project}/conflicts/{change}/resolveyard resolveyard_resolve
GET /v1/projects/{project}/agent-approvalsno command: a person's yard org agent-approvals reads the organization's setting through GET /api/orgs/{org}/agent-approvals; yard api GET /v1/projects/{project}/agent-approvals reads this routeyard_review
GET /v1/projects/{project}/review/pending/{pending}yard review approve --waitbrowser only: an approval of a protected change waits for the same person to confirm it in a browser signed in within 15 minutes, never with a key or a yard sign-in
POST /v1/projects/{project}/review/pending/{pending}browser only: an approval of a protected change waits for the same person to confirm it in a browser signed in within 15 minutes, never with a key or a yard sign-inbrowser only: an approval of a protected change waits for the same person to confirm it in a browser signed in within 15 minutes, never with a key or a yard sign-in
POST /v1/projects/{project}/review/{change}/{version}yard review approveyard_review

History, records and numbers

RouteyardMCP
GET /v1/projects/{project}/insightsyard insightsyard_insights
GET /v1/projects/{project}/valueyard valueyard_value
GET /v1/projects/{project}/whyyard whyyard_why
GET /v1/projects/{project}/attestations/{commit}yard verifyyard_attestation
GET /v1/projects/{project}/signeryard verify --signeryard_signer
GET /v1/projects/{project}/oplogyard logyard_log
POST /v1/projects/{project}/oplog/{seq}/undoyard undoyard_undo

Rules and checks

RouteyardMCP
PUT /v1/projects/{project}/rulesyard rules keepyard_rules_keep
POST /v1/projects/{project}/rules/uncheckedyard rules uncheckedyard_unchecked
POST /v1/projects/{project}/untested-guard/enforceyard rules enforceyard_untested_guard_enforce
GET /v1/projects/{project}/checks/triedyard check triedyard_checks_tried
GET /v1/projects/{project}/checks/logs/{log}yard check logyard_check_log
POST /v1/projects/{project}/checks/{check}/shellyard check shellyard_check_shell
DELETE /v1/projects/{project}/checks/{check}/shellyard check shell --doneyard_check_shell
PUT /v1/projects/{project}/checks/offyard check offyard_checks_off
POST /v1/projects/{project}/checks/tryyard check tryyard_checks_try
GET /api/projects/{project}/protected-areasyard rules protectsigned-in person: the change answers to a person's own access on GitHub, which an agent's key does not have
POST /api/projects/{project}/protected-areasyard rules protect --addsigned-in person: the change answers to a person's own access on GitHub, which an agent's key does not have

Settings and webhooks

RouteyardMCP
GET /v1/projects/{project}/settingsyard settingsyard_settings
PUT /v1/projects/{project}/checks/networkyard settings set check-networkyard_setting
GET /v1/projects/{project}/webhooksyard webhook listyard_webhooks
POST /v1/projects/{project}/webhooksyard webhook addyard_webhooks
DELETE /v1/projects/{project}/webhooks/{webhook}yard webhook removeyard_webhooks
POST /v1/projects/{project}/webhooks/{webhook}/testyard webhook testyard_webhooks
PUT /v1/projects/{project}/conflict-databrowser only: it is consent to use the project's conflict records: a person gives it in a browser signed in within 15 minutes, never with a key or a yard sign-in; an undo never changes itbrowser only: it is consent to use the project's conflict records: a person gives it in a browser signed in within 15 minutes, never with a key or a yard sign-in; an undo never changes it
PUT /v1/projects/{project}/wake-on-openyard settings set wake-on-openyard_setting
PUT /v1/projects/{project}/allowed-appsyard settings set allowed-appsyard_setting
PUT /v1/projects/{project}/self-approvalyard settings set self-approvalyard_setting
PUT /v1/projects/{project}/build-cacheyard settings set build-cacheyard_setting
PUT /v1/projects/{project}/grouped-mergeyard settings set grouped-mergeyard_setting
PUT /v1/projects/{project}/incremental-rebuildyard settings set incremental-rebuildyard_setting
PUT /v1/projects/{project}/merge-trainsyard settings set merge-trainsyard_setting
PUT /v1/projects/{project}/continue-mergedyard settings set continue-mergedyard_setting
PUT /v1/projects/{project}/checks-while-reviewingyard settings set checks-while-reviewingyard_setting
PUT /v1/projects/{project}/checks-before-submityard settings set checks-before-submityard_setting
PUT /v1/projects/{project}/engine-sizingyard settings set engine-sizingyard_setting
PUT /v1/projects/{project}/warm-houryard settings set warm-houryard_setting
PUT /v1/projects/{project}/task-permissionyard settings set who-submitsyard_setting
PUT /v1/projects/{project}/fixesyard settings set fixesyard_setting
PUT /v1/projects/{project}/docs-excusedyard settings set docs-excusedyard_setting
PUT /v1/projects/{project}/keep-failed-machinesyard settings set keep-failed-machinesyard_setting
PUT /v1/projects/{project}/reviewsyard settings set reviewsyard_setting

Keys

RouteyardMCP
POST /v1/projects/{project}/fleetsyard group createyard_agent_group_create
GET /v1/projects/{project}/fleetsyard group listyard_agent_groups
POST /v1/projects/{project}/admin-key/rotateyard project rotate-admin-keysigned-in person: a rotation stops every earlier admin key at once; Switchyard takes it only from a person signed in with admin access to the project, never from a key, the admin key itself included, so no MCP tool rotates it
POST /v1/projects/{project}/fleets/{fleet}/rotateyard groupyard_agent_group_key
POST /v1/projects/{project}/fleets/{fleet}/disableyard groupyard_agent_group_key
POST /v1/projects/{project}/fleets/{fleet}/enableyard groupyard_agent_group_key
POST /v1/projects/{project}/workspaces/{workspace}/keyyard branch keyyard_task_key
POST /v1/projects/{project}/fleets/offboardsigned-in person: it runs as part of removing a member from the organization, a person's change to its peoplesigned-in person: it runs as part of removing a member from the organization, a person's change to its people
POST /v1/projects/{project}/fleets/{fleet}/revokeyard groupyard_agent_group_key
GET /api/orgs/{org}/keysyard org keyssigned-in person: it acts for a person signed in, across projects; the MCP server answers inside one project, even for an agent signed in as its person over OAuth, so organization and account actions are not tools there
POST /api/orgs/{org}/keysyard org key createsigned-in person: it acts for a person signed in, across projects; the MCP server answers inside one project, even for an agent signed in as its person over OAuth, so organization and account actions are not tools there
POST /api/orgs/{org}/keys/{key}/{action}yard org keysigned-in person: it acts for a person signed in, across projects; the MCP server answers inside one project, even for an agent signed in as its person over OAuth, so organization and account actions are not tools there

The project

RouteyardMCP
POST /v1/projectsyard initsigned-in person: it acts for a person signed in, across projects; the MCP server answers inside one project, even for an agent signed in as its person over OAuth, so organization and account actions are not tools there
GET /v1/projects/{project}yard project showyard_whoami
GET /v1/projects/{project}/meyard whoamiyard_whoami
GET /v1/projects/{project}/creationno command: yard init waits on its own create and never reads it; yard api GET /v1/projects/{project}/creation reads itsigned-in person: it acts for a person signed in, across projects; the MCP server answers inside one project, even for an agent signed in as its person over OAuth, so organization and account actions are not tools there
GET /v1/projects/{project}/statusyard statusyard_status
GET /v1/projects/{project}/notificationssigned-in person: yard notifications reads your notifications of every project with your sign-in (GET /api/me/notifications); this route is the MCP sign-in's read inside one projectyard_notifications
GET /v1/projects/{project}/spending-limitno command: a person's yard org limit reads the organization's own limit through GET /api/orgs/{org}/billing/limit; yard api GET /v1/projects/{project}/spending-limit reads this route with the project's keyyard_spending_limit
GET /v1/projects/{project}/queue-healthyard queue healthyard_queue_health
PUT /v1/projects/{project}/pauseyard pauseyard_pause
DELETE /v1/projects/{project}/pauseyard resumeyard_resume
POST /v1/projects/{project}/copy-againyard copy-againyard_copy_again
POST /v1/projects/{project}/copy-historyyard copy-historyyard_copy_history
DELETE /v1/projects/{project}/copy-historyyard copy-history --cancelyard_copy_history
POST /v1/projects/{project}/engine/wakeyard engine wakeyard_wake_engine
POST /v1/projects/{project}/feedbackyard feedbackyard_feedback
POST /v1/projects/{project}/liveyard status --watchinternal: the live feed is a WebSocket held open for as long as one watches, which an MCP tool call cannot hold; MCP reads the same state with yard_status
POST /v1/projects/{project}/mcpinternal: it is the MCP server's own endpoint, which an MCP client calls; yard agent setup writes that client's configuration and calls nothinginternal: it is the MCP server's own endpoint, which every MCP tool goes through
GET /v1/projects/{project}/mcpinternal: it is the MCP server's own endpoint, which an MCP client calls; yard agent setup writes that client's configuration and calls nothinginternal: it is the MCP server's own endpoint, which every MCP tool goes through
DELETE /v1/projects/{project}/mcpinternal: it is the MCP server's own endpoint, which an MCP client calls; yard agent setup writes that client's configuration and calls nothinginternal: it is the MCP server's own endpoint, which every MCP tool goes through
GET /v1/projectsyard project listsigned-in person: it acts for a person signed in, across projects; the MCP server answers inside one project, even for an agent signed in as its person over OAuth, so organization and account actions are not tools there
GET /api/projects/{project}/deletionyard project deletionsigned-in person: it acts for a person signed in, across projects; the MCP server answers inside one project, even for an agent signed in as its person over OAuth, so organization and account actions are not tools there
POST /api/projects/{project}/deletionyard project deletesigned-in person: it acts for a person signed in, across projects; the MCP server answers inside one project, even for an agent signed in as its person over OAuth, so organization and account actions are not tools there
DELETE /api/projects/{project}/deletionyard project keepsigned-in person: it acts for a person signed in, across projects; the MCP server answers inside one project, even for an agent signed in as its person over OAuth, so organization and account actions are not tools there

Setting a repository up

RouteyardMCP
GET /api/githubinternal: the board reads the App's name here; yard setup github reads it with the installations from /api/onboarding/githubsigned-in person: it acts for a person signed in, across projects; the MCP server answers inside one project, even for an agent signed in as its person over OAuth, so organization and account actions are not tools there
GET /api/onboarding/githubyard setup githubsigned-in person: it acts for a person signed in, across projects; the MCP server answers inside one project, even for an agent signed in as its person over OAuth, so organization and account actions are not tools there
GET /api/onboarding/repositoriesyard setup repossigned-in person: it acts for a person signed in, across projects; the MCP server answers inside one project, even for an agent signed in as its person over OAuth, so organization and account actions are not tools there
GET /api/onboarding/analysisyard setup analyzesigned-in person: it acts for a person signed in, across projects; the MCP server answers inside one project, even for an agent signed in as its person over OAuth, so organization and account actions are not tools there
POST /api/onboarding/setupyard setup applysigned-in person: it answers to a person's own access on GitHub; it stays allowed with the person's yard sign-in, so an agent sets a repository up for its person, but no key or MCP tool does
POST /api/onboarding/checks-onyard setup checks-onsigned-in person: the change answers to a person's own access on GitHub, which an agent's key does not have

Organizations

RouteyardMCP
POST /api/auth/organization/createbrowser only: a person makes an organization in a browser where they are signed in, however long ago, so no agent makes one its person did not ask for; yard org create and an MCP sign-in are answered approval_required with the pagebrowser only: a person makes an organization in a browser where they are signed in, however long ago, so no agent makes one its person did not ask for; yard org create and an MCP sign-in are answered approval_required with the page
POST /api/auth/organization/updateyard org renamesigned-in person: it acts for a person signed in, across projects; the MCP server answers inside one project, even for an agent signed in as its person over OAuth, so organization and account actions are not tools there
POST /api/auth/organization/deletebrowser only: it cannot be undone, so an owner does it in a browser signed in within 15 minutes, never with a key or a yard sign-in; yard org delete prints that pagebrowser only: it cannot be undone, so an owner does it in a browser signed in within 15 minutes, never with a key or a yard sign-in; yard org delete prints that page
POST /api/auth/organization/accept-invitationyard org acceptsigned-in person: it changes the organization's people, teams or roles, which a person does themselves; no agent's key or tool may
POST /api/auth/organization/reject-invitationyard org declinesigned-in person: it changes the organization's people, teams or roles, which a person does themselves; no agent's key or tool may
POST /api/auth/organization/cancel-invitationyard org withdrawsigned-in person: it changes the organization's people, teams or roles, which a person does themselves; no agent's key or tool may
GET /api/orgsyard org listsigned-in person: it acts for a person signed in, across projects; the MCP server answers inside one project, even for an agent signed in as its person over OAuth, so organization and account actions are not tools there
GET /api/orgs/{org}yard org showsigned-in person: it acts for a person signed in, across projects; the MCP server answers inside one project, even for an agent signed in as its person over OAuth, so organization and account actions are not tools there
GET /api/teams/{org}no command: it is the older address of GET /api/orgs/{org}, which yard org show reads; yard api sends it with a person's sign-insigned-in person: it acts for a person signed in, across projects; the MCP server answers inside one project, even for an agent signed in as its person over OAuth, so organization and account actions are not tools there
GET /api/orgs/{org}/membersyard org memberssigned-in person: it acts for a person signed in, across projects; the MCP server answers inside one project, even for an agent signed in as its person over OAuth, so organization and account actions are not tools there
GET /api/orgs/{org}/invitationsyard org invitationssigned-in person: it acts for a person signed in, across projects; the MCP server answers inside one project, even for an agent signed in as its person over OAuth, so organization and account actions are not tools there
POST /api/invitationsyard org invitesigned-in person: it changes the organization's people, teams or roles, which a person does themselves; no agent's key or tool may
GET /api/invitationsyard org invitedsigned-in person: it acts for a person signed in, across projects; the MCP server answers inside one project, even for an agent signed in as its person over OAuth, so organization and account actions are not tools there
GET /api/orgs/{org}/members/{member}/keysyard org remove --keepsigned-in person: it changes the organization's people, teams or roles, which a person does themselves; no agent's key or tool may
POST /api/orgs/{org}/members/{member}/roleyard org rolesigned-in person: it changes the organization's people, teams or roles, which a person does themselves; no agent's key or tool may
POST /api/orgs/{org}/members/{member}/removeyard org removesigned-in person: it changes the organization's people, teams or roles, which a person does themselves; no agent's key or tool may
POST /api/orgs/{org}/leaveyard org leavesigned-in person: it changes the organization's people, teams or roles, which a person does themselves; no agent's key or tool may
GET /api/orgs/{org}/teamsyard org teamssigned-in person: it acts for a person signed in, across projects; the MCP server answers inside one project, even for an agent signed in as its person over OAuth, so organization and account actions are not tools there
POST /api/orgs/{org}/teamsyard org team createsigned-in person: it changes the organization's people, teams or roles, which a person does themselves; no agent's key or tool may
DELETE /api/orgs/{org}/teams/{team}yard org team deletesigned-in person: it changes the organization's people, teams or roles, which a person does themselves; no agent's key or tool may
GET /api/orgs/{org}/teams/{team}/membersyard org members --teamsigned-in person: it acts for a person signed in, across projects; the MCP server answers inside one project, even for an agent signed in as its person over OAuth, so organization and account actions are not tools there
POST /api/orgs/{org}/teams/{team}/membersbrowser only: who is on a team decides who reviews what, so a person adds someone to a team or takes them off in a browser signed in within 15 minutes; no key, yard sign-in or agent does it, and yard org team add and remove are answered approval_required with the teams pagesigned-in person: it changes the organization's people, teams or roles, which a person does themselves; no agent's key or tool may
DELETE /api/orgs/{org}/teams/{team}/members/{member}browser only: who is on a team decides who reviews what, so a person adds someone to a team or takes them off in a browser signed in within 15 minutes; no key, yard sign-in or agent does it, and yard org team add and remove are answered approval_required with the teams pagesigned-in person: it changes the organization's people, teams or roles, which a person does themselves; no agent's key or tool may
GET /api/orgs/{org}/ai-fixesyard org ai-fixessigned-in person: it acts for a person signed in, across projects; the MCP server answers inside one project, even for an agent signed in as its person over OAuth, so organization and account actions are not tools there
GET /api/orgs/{org}/agent-approvalsyard org agent-approvalssigned-in person: it acts for a person signed in, across projects; the MCP server answers inside one project, even for an agent signed in as its person over OAuth, so organization and account actions are not tools there
PUT /api/orgs/{org}/agent-approvalsbrowser only: letting agents approve work is changed only by an owner or admin in a browser signed in within 15 minutes, so no agent turns it on for itself; yard org agent-approvals on prints that pagebrowser only: letting agents approve work is changed only by an owner or admin in a browser signed in within 15 minutes, so no agent turns it on for itself; yard org agent-approvals on prints that page
GET /api/orgs/{org}/agent-policysigned-in person: read on the organization's team page; no yard command reads it yetsigned-in person: it acts for a person signed in, across projects; the MCP server answers inside one project, even for an agent signed in as its person over OAuth, so organization and account actions are not tools there
PUT /api/orgs/{org}/agent-policybrowser only: what agents may do in an organization is changed only by an owner or admin in a browser signed in within 15 minutes, so no agent turns its own limits off; yard and MCP sign-ins are answered approval_required with the team pagebrowser only: what agents may do in an organization is changed only by an owner or admin in a browser signed in within 15 minutes, so no agent turns its own limits off; yard and MCP sign-ins are answered approval_required with the team page
PUT /api/orgs/{org}/ai-fixesbrowser only: AI fixes spend on the organization's bill, so a person turns them on or off in a browser signed in within 15 minutes; yard org ai-fixes on prints that pagebrowser only: AI fixes spend on the organization's bill, so a person turns them on or off in a browser signed in within 15 minutes; yard org ai-fixes on prints that page
GET /api/orgs/{org}/engine-sizingyard org engine-sizingsigned-in person: it acts for a person signed in, across projects; the MCP server answers inside one project, even for an agent signed in as its person over OAuth, so organization and account actions are not tools there
PUT /api/orgs/{org}/engine-sizingyard org engine-sizingsigned-in person: it acts for a person signed in, across projects; the MCP server answers inside one project, even for an agent signed in as its person over OAuth, so organization and account actions are not tools there
GET /api/orgs/{org}/valueyard org valuesigned-in person: it acts for a person signed in, across projects; the MCP server answers inside one project, even for an agent signed in as its person over OAuth, so organization and account actions are not tools there
GET /api/orgs/{org}/usageyard org usagesigned-in person: it acts for a person signed in, across projects; the MCP server answers inside one project, even for an agent signed in as its person over OAuth, so organization and account actions are not tools there
GET /api/orgs/{org}/billingyard org billingsigned-in person: it acts for a person signed in, across projects; the MCP server answers inside one project, even for an agent signed in as its person over OAuth, so organization and account actions are not tools there
GET /api/orgs/{org}/billing/runs.csvyard org billing --csvsigned-in person: it acts for a person signed in, across projects; the MCP server answers inside one project, even for an agent signed in as its person over OAuth, so organization and account actions are not tools there
PUT /api/orgs/{org}/billing/contactsyard org billing contactsbrowser only: it changes the organization's billing, which a person who looks after billing does on the billing page, in a browser signed in within 15 minutes (yard org billing contacts prints that page); no agent's key or tool may
GET /api/orgs/{org}/billing/limityard org limitsigned-in person: it acts for a person signed in, across projects; the MCP server answers inside one project, even for an agent signed in as its person over OAuth, so organization and account actions are not tools there
PUT /api/orgs/{org}/billing/limitbrowser only: a spending limit is set, raised or removed only on the billing page by a person signed in in a browser in the last 15 minutes, never with a key or a yard sign-in; yard org limit set prints the link that confirms itbrowser only: a spending limit is set, raised or removed only on the billing page by a person signed in in a browser in the last 15 minutes, never with a key or a yard sign-in; yard org limit set prints the link that confirms it
POST /api/orgs/{org}/billing/limit/raisebrowser only: a spending limit is set, raised or removed only on the billing page by a person signed in in a browser in the last 15 minutes, never with a key or a yard sign-in; yard org limit set prints the link that confirms itbrowser only: a spending limit is set, raised or removed only on the billing page by a person signed in in a browser in the last 15 minutes, never with a key or a yard sign-in; yard org limit set prints the link that confirms it
GET /api/orgs/{org}/billing/cardyard org cardsigned-in person: it shows the card and the bill's amounts, which only the people who look after billing read; agents never read the spend, and a payment hold reaches them through yard_status and yard_spending_limit
POST /api/orgs/{org}/billing/cardyard org card addsigned-in person: a person who looks after billing enters the card on Stripe's own page; no key opens it, yard is answered approval_required with the billing page, and the MCP server works inside one project with that project's keys
GET /api/orgs/{org}/billing/invoicesyard org invoicessigned-in person: it shows the card and the bill's amounts, which only the people who look after billing read; agents never read the spend, and a payment hold reaches them through yard_status and yard_spending_limit
GET /api/orgs/{org}/billing/limit/offersigned-in person: it reads what a billing email's raise link offers, for the billing page, where the raise itself takes a browsersigned-in person: it reads what a billing email's raise link offers, for the billing page, where the raise itself takes a browser
GET /api/orgs/{org}/audityard org auditsigned-in person: it acts for a person signed in, across projects; the MCP server answers inside one project, even for an agent signed in as its person over OAuth, so organization and account actions are not tools there
GET /api/orgs/{org}/deleted-projectsyard org deletedsigned-in person: it acts for a person signed in, across projects; the MCP server answers inside one project, even for an agent signed in as its person over OAuth, so organization and account actions are not tools there
GET /api/orgs/{org}/ssoyard org ssosigned-in person: it acts for a person signed in, across projects; the MCP server answers inside one project, even for an agent signed in as its person over OAuth, so organization and account actions are not tools there
POST /api/orgs/{org}/sso/{provider}/verifyyard org sso verifysigned-in person: it acts for a person signed in, across projects; the MCP server answers inside one project, even for an agent signed in as its person over OAuth, so organization and account actions are not tools there
GET /api/orgs/{org}/scimyard org scimsigned-in person: it acts for a person signed in, across projects; the MCP server answers inside one project, even for an agent signed in as its person over OAuth, so organization and account actions are not tools there
GET /api/orgs/{org}/ip-allowlistsigned-in person: a member with access to it reads it with their own sign-in, which yard holds; yard has no command for it, and yard api sends it with that sign-insigned-in person: it acts for a person signed in, across projects; the MCP server answers inside one project, even for an agent signed in as its person over OAuth, so organization and account actions are not tools there
GET /api/orgs/{org}/two-factoryard org two-factorsigned-in person: it acts for a person signed in, across projects; the MCP server answers inside one project, even for an agent signed in as its person over OAuth, so organization and account actions are not tools there

Your account

RouteyardMCP
POST /api/auth/device/codeyard login --githubinternal: it is how yard login --github signs in; an MCP client signs in its own way (Claude Code and Codex) or sends a key
POST /api/auth/device/tokenyard login --githubinternal: it is how yard login --github signs in; an MCP client signs in its own way (Claude Code and Codex) or sends a key
POST /api/auth/sign-outyard logoutsigned-in person: it acts for a person signed in, across projects; the MCP server answers inside one project, even for an agent signed in as its person over OAuth, so organization and account actions are not tools there
POST /api/auth/revoke-sessionsyard account sessions end --allsigned-in person: it acts for a person signed in, across projects; the MCP server answers inside one project, even for an agent signed in as its person over OAuth, so organization and account actions are not tools there
POST /api/auth/delete-userbrowser only: it cannot be undone, so a person does it in a browser signed in within 15 minutes, never with a yard sign-inbrowser only: it cannot be undone, so a person does it in a browser signed in within 15 minutes, never with a yard sign-in
GET /api/meyard whoamisigned-in person: it acts for a person signed in, across projects; the MCP server answers inside one project, even for an agent signed in as its person over OAuth, so organization and account actions are not tools there
GET /api/me/emailsyard account emailssigned-in person: it acts for a person signed in, across projects; the MCP server answers inside one project, even for an agent signed in as its person over OAuth, so organization and account actions are not tools there
PUT /api/me/emailsyard account emails setsigned-in person: it acts for a person signed in, across projects; the MCP server answers inside one project, even for an agent signed in as its person over OAuth, so organization and account actions are not tools there
GET /api/me/notificationsyard notificationssigned-in person: a person's own notifications, read and changed signed in as that person; an agent's MCP sign-in reads only its person's open reviews and tasks of one project, with yard_notifications (GET /v1/projects/{project}/notifications)
POST /api/me/notifications/seensigned-in person: a browser marks what the bell showed; yard notifications marks nothing seen, and yard notifications read marks rows readsigned-in person: a person's own notifications, read and changed signed in as that person; an agent's MCP sign-in reads only its person's open reviews and tasks of one project, with yard_notifications (GET /v1/projects/{project}/notifications)
POST /api/me/notifications/readyard notifications readsigned-in person: a person's own notifications, read and changed signed in as that person; an agent's MCP sign-in reads only its person's open reviews and tasks of one project, with yard_notifications (GET /v1/projects/{project}/notifications)
POST /api/me/notifications/clearyard notifications clearsigned-in person: a person's own notifications, read and changed signed in as that person; an agent's MCP sign-in reads only its person's open reviews and tasks of one project, with yard_notifications (GET /v1/projects/{project}/notifications)
POST /api/me/notifications/muteyard notifications mutesigned-in person: a person's own notifications, read and changed signed in as that person; an agent's MCP sign-in reads only its person's open reviews and tasks of one project, with yard_notifications (GET /v1/projects/{project}/notifications)
GET /api/me/notification-settingsyard notifications settingssigned-in person: a person's own notifications, read and changed signed in as that person; an agent's MCP sign-in reads only its person's open reviews and tasks of one project, with yard_notifications (GET /v1/projects/{project}/notifications)
PUT /api/me/notification-settingsyard notifications settings setsigned-in person: a person's own notifications, read and changed signed in as that person; an agent's MCP sign-in reads only its person's open reviews and tasks of one project, with yard_notifications (GET /v1/projects/{project}/notifications)
POST /api/me/liveinternal: the bell's live connection is a WebSocket held open while a page shows the bell, which a yard command or an MCP tool call does not holdinternal: the bell's live connection is a WebSocket held open while a page shows the bell, which a yard command or an MCP tool call does not hold
GET /api/me/two-factorbrowser only: a person confirms it in a browser, signed in with GitHub in the last 15 minutes; no key and no agent confirms itsigned-in person: it acts for a person signed in, across projects; the MCP server answers inside one project, even for an agent signed in as its person over OAuth, so organization and account actions are not tools there
GET /api/me/agentsyard account agentssigned-in person: it acts for a person signed in, across projects; the MCP server answers inside one project, even for an agent signed in as its person over OAuth, so organization and account actions are not tools there
DELETE /api/me/agents/{agent}yard account agents endsigned-in person: it acts for a person signed in, across projects; the MCP server answers inside one project, even for an agent signed in as its person over OAuth, so organization and account actions are not tools there
GET /api/sessionsyard account sessionssigned-in person: it acts for a person signed in, across projects; the MCP server answers inside one project, even for an agent signed in as its person over OAuth, so organization and account actions are not tools there
DELETE /api/sessions/{session}yard account sessions endsigned-in person: it acts for a person signed in, across projects; the MCP server answers inside one project, even for an agent signed in as its person over OAuth, so organization and account actions are not tools there

The service

RouteyardMCP
GET /v1yard doctorinternal: a liveness answer for probes and yard doctor; an MCP session is already connected
GET /v1/upstreamyard statusyard_status
GET /v1/openapi.jsoninternal: it describes the API for machines; yard and the MCP server are two of the ways it describesinternal: it describes the API for machines; yard and the MCP server are two of the ways it describes

<!-- end of every route -->