AgileHero

AI agent access · MCP server built in

AI agents for project managementThe whole product, agent-operable.

AgileHero ships a built-in MCP server: hand your AI assistant one personal token and it can work your boards, whiteboards, retros, roadmap, and wiki — 49 tools, always inside your permissions. Read the tool reference and the AgileHero Markup spec.

Start free On every plan, including Free · 500 agent calls/day · 5,000 on Pro

Connected in three steps

1

Create a token for your client

Settings → MCP Server: pick the client (Claude Code, Cursor, VS Code, Codex, and more), name the token, and get its config with the token already filled in — shown once, stored only as a hash. One token per client, revocable one at a time; creating and revoking ask you to re-authenticate first.

2

Connect your assistant

Point any MCP-compatible client at https://api.agilehero.io/mcp with the header Authorization: Bearer <token> — or paste the ready-made config. Per-client setup guides and the full tool reference live at agilehero.io/docs/mcp, generated from the server itself.

3

Put it to work

49 tools cover more or less everything you can do in the UI: cards, epics, and comments; whiteboards and diagrams; retrospectives; roadmap slots; wiki pages; delivery metrics; and full-text search across it all.

An agent-operable product

The whole product, operable

Agents don't stop at the board: they can draw on whiteboards, add retro items, schedule roadmap slots, write wiki pages, and read your delivery metrics — the same surfaces you use.

Your permissions, exactly

Every tool authorizes through the same rules as the human UI. A resource you can't access looks identical to one that doesn't exist — no probing — and the whole surface fails closed.

AgileHero Markup, built for agents

Rich text is structured content, so agents write AHM — a strict XML markup that maps exactly to what the editor renders. No lossy markdown round-trips, and agents fetch the spec with a single tool call.

Two-step edits by design

Rich-text updates preview first: the agent sees the consequences of its edit and gets a token, then confirms. Descriptions and wiki pages never get clobbered blind.

Agents can draw

Stickies, shapes, connectors, frames — even whole diagrams laid out server-side — and whiteboard elements convert into real cards and epics, agent-driven end to end.

One token per client, all yours

Every token acts as you and only you. Revoke the Cursor token without touching Claude Code, see when each one was last used, and re-authenticate before you create or revoke one.

Not just cards on a board

An agent that can only file cards is a chatbot with a form. Connected over MCP, yours can draw real diagrams on the whiteboard — laid out server-side — write and update wiki pages in AgileHero Markup so the formatting renders exactly, queue talking points for the next retrospective, schedule epics on the roadmap, and read your delivery metrics before standup. Same token, same permissions, all 49 tools.

What can an AI agent do in a project tracker?

Most trackers give AI a read-only API and call it an integration: the assistant can summarize your backlog but can't touch it. The useful version is an agent that operates the product — files the bug as a real card with an estimate, sketches the architecture on a whiteboard and converts it into epics, drafts the release notes in the wiki, and checks the metrics before standup — while staying inside the permission system your team already trusts.

AgileHero ships that as a built-in MCP server: 49 tools spanning cards, epics, whiteboards, retrospectives, the roadmap, wiki pages, metrics, and search — every call authorized as you, in the projects you can see, and nothing more. Rich text uses AgileHero Markup, a strict XML markup built for agents (most tools bolt extensions onto markdown; we chose fidelity over familiarity, so an agent's edit renders exactly as intended), and rich-text updates are two-step by design: preview the consequences, then confirm.

Common questions

Is MCP access free?

Yes — the MCP server is available on every plan, including Free, with a daily fair-use quota: 500 agent calls per user per day on Free workspaces, 5,000 on Pro (trials count as Pro). The quota is per user per workspace and resets at 00:00 UTC; reading the AHM spec and discovering your projects never count against it.

Which AI tools can connect?

Claude — and any client that speaks MCP (Model Context Protocol). The server exposes one endpoint; the client authenticates with your personal token as a bearer credential. The complete tool reference is public at agilehero.io/docs/mcp.

Is it safe to let an agent write to the board?

The agent inherits exactly your permissions: the same authorization rules as the UI, project by project, fail-closed. A resource you can't access is reported as if it doesn't exist, the token is stored only as a hash, you can revoke it at any moment — and rich-text edits are two-step, so the agent previews the consequences before anything is overwritten.

What exactly can an agent do?

49 tools across the product: discover and read boards, cards, epics, labels, members, whiteboards, retrospectives, roadmap slots, wiki pages, and delivery metrics; full-text search across everything; create, update, and move cards; author epics; place and reschedule roadmap slots; add and move retro items; draw on whiteboards and convert elements into work; and write wiki pages.

Can agents use the whiteboard?

Yes — agents can create whiteboards, place stickies, shapes, connectors, and frames, generate whole diagrams with server-side layout, review a whiteboard's contents, and convert elements or entire mind maps into real cards and epics.

Why AgileHero Markup instead of markdown?

Rich text in AgileHero is structured editor content, and markdown-with-extensions loses information in both directions. AHM is a strict XML markup that maps one-to-one to what the editor renders — an agent fetches the spec with a single tool call and writes formatting that never gets mangled. The specification is public at agilehero.io/docs/mcp/agilehero-markup.

What does an observer's agent get?

Read-only, same as the observer: the token carries your role, so an agent connected with an observer's token can look but never write.

Start with two seats, free forever

Every plan begins as a 14-day trial of every premium module. Set up your first project in minutes.

Try it for free

14-day free trial · no credit card required