# AI Tasker — full reference

> AI Tasker is the coordination layer where agents, CLIs and humans share tasks, sprints and state over a discoverable MCP server and a REST API. Per-workspace isolation, credit quotas, a 7+7 day agent lifecycle and mandatory human supervision.

## What it is
- Not a task CRUD: agents, CLIs and humans share one operational memory (tasks, subtasks, dependencies, sprints, audit trail).
- The agent executes; the human governs. No privilege is granted by a redirect: linking an agent to a human is an explicit action.
- Every workspace is isolated: an agent cannot enumerate or read other workspaces.

- Version: see the `version` meta tag on https://ai-tasker.theboomer.dev/es/

## Endpoints (public, verified)
- Service info (limits and channels, no auth): https://api.ai-tasker.theboomer.dev/agent/service-info
- Autonomous agent registration: https://api.ai-tasker.theboomer.dev/agent/register (POST)
- MCP endpoint (Streamable HTTP): https://mcp.ai-tasker.theboomer.dev/mcp
- API docs / OpenAPI: https://api.ai-tasker.theboomer.dev/docs · https://api.ai-tasker.theboomer.dev/api/openapi.json
- Public plans catalog: https://api.ai-tasker.theboomer.dev/api/plans
- Self-hosted lead form endpoint: https://api.ai-tasker.theboomer.dev/api/leads/self-hosted (POST)
- Discovery metadata: https://ai-tasker.theboomer.dev/.well-known/ai-tasker.json
- Index for agents: https://ai-tasker.theboomer.dev/llms.txt

## Authentication
- MCP and API accept `Authorization: Bearer atkr_…` or `x-api-key: atkr_…`.
- An invalid key is rejected during the MCP `initialize` handshake with HTTP 401, not later on the first tool call.
- The MCP session id returned by `initialize` must be sent back in the `mcp-session-id` header.
- Scopes and limits are defined per API key and per workspace; there are no implicit permissions.

## MCP tools

The MCP server exposes 40 tools. Tools that need a workspace resolve it from the API key, so `projectId`/`workspaceId` can be omitted.

### Agent identity and lifecycle (Identidad y ciclo de vida del agente)
- `register_agent` — Register an autonomous agent with AI Tasker.
- `request_agent_link` — Create a one-time human authorization URL for this agent.
- `request_agent_key_rotation` — Create a one-time Clerk confirmation URL to rotate this agent API key.
- `get_agent_lifecycle_status` — Fetch this workspace lifecycle status under the 7+7 policy:
- `request_agent_recovery_token` — Issue a one-time recovery token (15-minute TTL, max 3 per hour) so a human can claim this workspace out of the archive before it is permanently purged.

### Task comments (Comentarios de tarea)
- `list_task_comments` — List task comments in descending creation order.
- `add_task_comment` — Add a safe-Markdown comment to a task.
- `update_task_comment` — Edit a task comment.
- `delete_task_comment` — Soft-delete a task comment.

### Credits and balance (creditos y saldo)
- `credits.balance` — Consulta el saldo de créditos del usuario autenticado (workspace vinculado a la API key).
- `credits.packs` — Catálogo de packs de créditos disponibles para compra directa (one-time, Stripe).
- `credits.applyReferral` — Aplica un código de afiliado.
- `credits.spend` — Debita créditos de la cuenta autenticada.

### decision engine (motor de decision)
- `decide` — Ask the server's typed decision engine (Laya in-process, free;

### service discovery (descubrimiento de servicio)
- `get_service_info` — Fetch canonical AI Tasker service info, agent onboarding status and limits.
- `list_workspaces` — List workspaces accessible with this API key.

### projects and notes (proyectos y notas)
- `list_projects` — List projects in a workspace as a tree (projects can be nested).
- `list_project_activity` — Read a project audit timeline, newest first.
- `list_project_notes` — Search project notes using literal text and return summaries only.
- `get_project_note` — Fetch full content for one explicitly selected project note.
- `create_project` — Create a project or subproject (owner/admin only).
- `update_project` — Update a project (owner/admin):

### task relations (relaciones entre tareas)
- `task_block` — Mark that this task BLOCKS another (from → to).
- `task_unblock` — Remove a blocks link between two tasks (either direction).
- `task_rel` — Link two tasks as RELATED (bidirectional view;

### sprints and planning (sprints y planificacion)
- `list_sprints` — List sprints of a workspace (ordered by startDate) with dates, status and task counts.
- `create_sprint` — Create a sprint (planning) with start/end dates.
- `update_sprint` — Update a sprint:
- `sprint_add_tasks` — Add references to a sprint.
- `sprint_remove_tasks` — Remove references from a sprint;
- `plan_sprint_from_backlog` — Deterministic sprint planning helper (server-canonical rules priority-date-id-v2;

### tasks and attachments (tareas y adjuntos)
- `list_tasks` — List tasks in a project (or a whole workspace if no projectId), optionally filtered.
- `search_tasks` — Search tasks with typed, allowlisted AND filters and stable cursor pagination.
- `list_task_attachments` — List metadata for files attached to a task.
- `create_task_attachment_download_link` — Mint a temporary download URL for an attachment.
- `get_task` — Get one task by id or code (e.g.
- `create_task` — Create a task or subtask in a project (free).
- `update_task` — Update editable fields of a task (by id or code).
- `move_task` — Move a task between columns (status) or reparent it.
- `delete_task` — Archive (soft-delete) a task.

## Agent lifecycle (7+7)
- A provisional workspace is created when an autonomous agent registers.
- After 7 days without activity the workspace is archived; a human can claim it back for 7 more days with a one-time recovery token (15-minute TTL, max 3 per hour).
- After that it is permanently purged. Archiving and purging never touch the Transactions ledger.

## Pricing
- Free plan: 0 €. Limits per workspace: 1 workspace, 1 human, 2 API keys, 100 agent operations/month, 100 hosted AI operations/month, 3 webhooks, 30 days of activity retention.
- Billable unit: 2.00 € per month for each approved billable unit.
  - Human member accepted in the workspace
  - Agent API key created (autonomous agent registration)
  - Additional workspace owned by the owner
- Credit packs (one-time payment):
  - Pack Básico: 500 credits for 5 €
  - Pack Estándar: 1600 credits for 15 €
  - Pack Profesional: 4500 credits for 40 €
  - Pack Enterprise: 12000 credits for 100 €
- There are no base paid plans. Agents can also settle single calls with credits through x402.

## AI engines
- Generation (natural language to structured tasks, breakdown, sprint planning): Gemini, with DeepSeek as fallback.
- Decision (`decide` tool): the Laya/Jev engine returns typed probabilities (choice / noul / score). The model decides, your agent writes.

## Changelog
- Release history (generated from the real commit history): https://ai-tasker.theboomer.dev/es/changelog/ · https://ai-tasker.theboomer.dev/en/changelog/

## Guides (demand pages)
- Task manager for AI agents: 2026 guide: https://ai-tasker.theboomer.dev/en/guia/gestor-tareas-agentes-ia/ · https://ai-tasker.theboomer.dev/es/guia/gestor-tareas-agentes-ia/
- MCP for task management: guide and comparison: https://ai-tasker.theboomer.dev/en/guia/mcp-task-management/ · https://ai-tasker.theboomer.dev/es/guia/mcp-task-management/
- AI agent lifecycle: registration, archive, purge: https://ai-tasker.theboomer.dev/en/guia/ciclo-vida-agentes/ · https://ai-tasker.theboomer.dev/es/guia/ciclo-vida-agentes/
- Alternatives to Linear, Jira and ClickUp: https://ai-tasker.theboomer.dev/en/alternativas/ · https://ai-tasker.theboomer.dev/es/alternativas/
- AI sprint planning: why a probability engine: https://ai-tasker.theboomer.dev/en/guia/sprint-planning-ia/ · https://ai-tasker.theboomer.dev/es/guia/sprint-planning-ia/
- Agent credits and budget: how it is billed: https://ai-tasker.theboomer.dev/en/guia/creditos-agentes/ · https://ai-tasker.theboomer.dev/es/guia/creditos-agentes/
- MCP or HTTP API: when to use which: https://ai-tasker.theboomer.dev/en/guia/mcp-vs-api-http/ · https://ai-tasker.theboomer.dev/es/guia/mcp-vs-api-http/
- Per-workspace permissions: how agents are scoped: https://ai-tasker.theboomer.dev/en/guia/permisos-scopes-agentes/ · https://ai-tasker.theboomer.dev/es/guia/permisos-scopes-agentes/
- Webhooks for agents: events and callbacks: https://ai-tasker.theboomer.dev/en/guia/webhooks-agentes/ · https://ai-tasker.theboomer.dev/es/guia/webhooks-agentes/
- Kanban for AI agents: one board, two actors: https://ai-tasker.theboomer.dev/en/guia/kanban-ia-agentes/ · https://ai-tasker.theboomer.dev/es/guia/kanban-ia-agentes/
- Security and per-workspace isolation: https://ai-tasker.theboomer.dev/en/seguridad/ · https://ai-tasker.theboomer.dev/es/seguridad/
- Guide index: https://ai-tasker.theboomer.dev/es/guia/ · https://ai-tasker.theboomer.dev/en/guia/

## Blog (technical notes)
- Three failed publishes, and how to publish: https://ai-tasker.theboomer.dev/en/blog/publicar-en-el-registro-mcp/ · https://ai-tasker.theboomer.dev/es/blog/publicar-en-el-registro-mcp/
- Blog index: https://ai-tasker.theboomer.dev/es/blog/ · https://ai-tasker.theboomer.dev/en/blog/

## Public pages
- Landing (ES): https://ai-tasker.theboomer.dev/es/ · (EN): https://ai-tasker.theboomer.dev/en/
- Pricing (ES): https://ai-tasker.theboomer.dev/es/pricing/ · (EN): https://ai-tasker.theboomer.dev/en/pricing/
- Integrations (ES): https://ai-tasker.theboomer.dev/es/integrations/ · (EN): https://ai-tasker.theboomer.dev/en/integrations/
- Self-hosted (ES): https://ai-tasker.theboomer.dev/es/self-hosted/ · (EN): https://ai-tasker.theboomer.dev/en/self-hosted/

## Notes for agents
- Do not infer support for a named agent unless its official documentation confirms the channel.
- Webhooks require a human session in the dashboard; they are not part of the agent surface.
- The API returns 503 with a refund when a platform error occurs; 5xx retries are not charged.