← Docs home

MCP reference / Tools

Tools

The 173 MCP tools TAM exposes to Claude, Cursor, and other MCP clients: what each domain lets an AI agent actually do, plus naming, scopes, and rate limits.

TAM's MCP server has 173 tools in its catalog. What tools/list actually returns to your client (Claude, Cursor, Codex, or your own agent) is narrower: only the tools the connection's grant allows. A personal connection typically sees everything your own account can do; a scoped service account sees only the tools covered by the scopes it was set up with. Every tool maps to a real backend operation, so an AI agent can only do what the token it's running as is actually allowed to do, not just what's hidden from the list.

How tools are named

When your client asks for the tool list, the names it actually gets back use hyphens: tam-issue-list, tam-wiki-create-page, tam-sprint-update. That's how you'll see them in Claude, and it's the name you call. (Internally TAM's registry uses dots, tam.issue.list, but Claude and most MCP clients reject dots in a tool name, so the server converts every name at the door.) Alongside those 159 converted names, 14 tools also ship under short, dot-free aliases for older clients: create_issue, search, fetch, wiki_search, and similar. Both forms of a name call the same backend operation.

What a scope actually restricts

Every tool requires specific scopes to call it (for example, creating an issue requires issue.create). A connection's token carries a specific scope list agreed at setup, an optional project allowlist, and sometimes an explicit tool allowlist on top of that. If a tool isn't in the token's allowed list, the call is rejected before it reaches TAM's backend, not just hidden from the tool listing. That's the mechanism behind service accounts: an agent wired up for sprint planning can get issue.read and sprint.write without ever getting project.delete or team.member.remove, regardless of what the model decides to try.

What you can do with them

DomainWhat an agent can doExample tool
Issues (25 tools)Create, triage, assign, comment on, and move tasks through your workflow; claim work, link related issues, track progress on a task end to endtam-issue-create, tam-issue-claim
Projects (20 tools)Create and configure projects, read kanban/sprint/feature boards and dashboards, manage who's on a projecttam-project-list, tam-project-dashboard-get
Wiki (15 tools)Search, read, write, and export documentation and knowledge-base pages as Markdowntam-wiki-search, tam-wiki-create-page
Workspace (16 tools)Manage members, roles, and invitations; read and update workspace settingstam-workspace-get, tam-workspace-member-invite
Teams (13 tools)Organize members into squads, guilds, or chapters; add or remove people in bulktam-team-list, tam-team-member-add-bulk
Sprints & workflow (11 tools)Plan and run sprints, move issues through custom workflow statuses, read the backlogtam-sprint-create, tam-workflow-transition
Integrations & GitHub (12 tools)Connect GitHub, import issues, sync commits, and check PR status against a tasktam-integration-configure, tam-github-pr-status
Custom fields & labels (10 tools)Define custom fields and labels for your workflow, then tag and filter issues by themtam-custom-field-definition-create, tam-label-list
Time tracking (5 tools)Log time against a task and pull time reportstam-time-entry-create, tam-time-entry-list
Automation rules (4 tools)Create rules that auto-assign, auto-move, or auto-label issues on a triggertam-automation-rule-create
Analytics & reporting (4 tools)Pull burndown charts, project health snapshots, and generate a weekly status reporttam-analytics-burndown-get, tam-report-weekly-generate
Bots & service accounts (5 tools)Create, inspect, and revoke the service accounts other AI agents connect throughtam-bot-list, tam-bot-status-get
Everything else (33 tools)Billing plans, usage rollups, notifications, file attachments, your own profile, the audit log, saved views, and searchtam-notification-list, tam-attachment-upload-prepare

This table maps the full 173-tool catalog; tools/list on your own connection returns whichever subset your grant actually authorizes, so don't be surprised if you see fewer. That's expected access control working as intended, not a missing feature.

Read tools are rate-limited, writes are idempotent where it matters

Read tools are rate-limited per connection to keep the service responsive for everyone, so a well-behaved client backs off and retries rather than hammering the endpoint. Mutating tools that create or move state, like issue creation or sprint transitions, carry idempotency guarantees so a client retrying after a timeout doesn't double-create the same issue. If you're building a client that retries automatically, pass the same idempotency key on retry rather than generating a new one.

Calling a tool

Server URL: Published when the global gateway launches

Once connected, ask your agent to list issues, create a task, or search the wiki; the client discovers every tool the current token is scoped for through the standard MCP tools/list call. No separate API key or SDK integration needed beyond the MCP connection itself.

See MCP Resources for the read-only context objects tools operate on, and MCP Prompts for the packaged workflows built on top of these tools.