agentda

Agentda documentation

Agentda has no task UI — your agent is the interface. This page explains connecting and the core concepts; everything else happens in conversation.

Connect

claude mcp add --transport http agentda https://myagentda.com/api/mcp

The OAuth flow opens on first access (sign-in via magic link). Other MCP clients: endpoint https://myagentda.com/api/mcp (Streamable HTTP, OAuth with PKCE and Dynamic Client Registration).

Using Smithery? Agentda is also listed on Smithery (smithery mcp add maurer/agentda) — connections then run through Smithery's gateway; the direct endpoint above stays the recommended path.

Concepts

  • Workspace = domain. Sign up with your work email; colleagues on the same domain join automatically (admins can turn this off).
  • Tasks are called A-1, A-17, … Statuses: open, in progress, waiting, done. Due date and follow-up (snooze) are separate concepts.
  • Delegation always requires a reason; the recipient accepts, declines, or hands it back. Delegating to any email address turns external people into guests: their own task page via link, no account, free.
  • History: every task carries an immutable event log — your agent answers status questions from it, no need to ask around.
  • Visibility through participation: owner, assignee, CC, explicitly shared, guests. No visibility by hierarchy.
  • Reply-to-task: replies to notification emails are saved as comments on the task.
  • Attachments: up to 4 MB per file, uploaded via one-time URLs (your agent handles this with curl).
  • Topics are called T-1, T-2, … — shared knowledge scopes holding markdown documents: the shared memory of every agent on the team (research, decisions, how-tos). Private by default; the creator names writers and readers, or makes a topic workspace-readable (write access stays with the creator and named writers). Tasks can be linked to topics — which shares the task with everyone who can read the topic.
  • Response forms & checklists: when a task expects a structured reply (poll, RSVP, meal plan), the agent attaches a form (yes/no, choices, date, number, free text). Guests answer with real form controls on the task page, agents answer conversationally; the latest answer per person counts, and all answers are collected on the task. Checkbox markers (☐/☑ or “- [ ]”) in descriptions become tickable checkboxes for guests.

Pricing, trial, export

Free forever for solo users. As soon as a second member joins, the workspace becomes a team: 30 days free, then €5/user/month (billed yearly). When a team trial expires the workspace becomes read-only until a subscription is active — reading, searching, and the full export (https://myagentda.com/api/export, signed in) always remain free; solo workspaces stay free and writable. Guests are always free.

Agentda in your chat

Task updates can flow into your chat as a room message — via outgoing webhook (Slack, Zulip, Discord, Mattermost, Buzz …). Set it up in conversation with your agent: “Post task updates to our Slack channel” — the agent callsadmin_tenant with set_webhook and the incoming-webhook URL (admins only, max. 3, https only). Every delivery is HMAC-signed (X-Agentda-Signature: SHA-256 over <t>.<body> using the secret shown once at setup). Slack URLs automatically get the{ text } format.

In agentic workspaces (Buzz, Kollab & co.) it’s even more direct: their agents speak MCP — they simply connect to the Agentda endpoint themselves and work with tasks, delegations, and history like any other agent. Step-by-step answers per tool (Slack, Teams, Zulip, Discord …) in the integrations FAQ.

Machine-readable