Glossary of terms — AI Tasker

This glossary exists for a practical reason: most false claims about AI Tasker start with a misunderstood term. Here every word means what it means in the code or on the public surface, in two or three sentences, with no creative synonyms. If a term is missing and you need it, the rule is not to extrapolate it: look it up in the full reference and, if it is not there either, do not assert it.

Product and organisation

Term	Definition
Workspace	Unit of consumption, quota and isolation. Each workspace has its own members, API keys, projects and monthly allowance. A workspace never sees another workspace’s data.
Project	Grouping of tasks and sprints inside a workspace. It is organisation, not perimeter: a credential that reaches the workspace sees all of its projects.
Sprint	Planning window with its associated tasks. It can be generated with hosted AI, and that generation is an operation that consumes quota.
Task	The unit of work shared by humans and agents. It has a readable code (`TSK-42`) in addition to an internal id, and the code requires a scope (project or workspace) to resolve.
Actor	Who performs an action: a person or an agent. The channel (`human` or `agent`) is recorded in the activity and in the operations.
Linked human	The person who claims a workspace created by an autonomous agent. From then on the agent identity becomes `linked` and its provisional counters stop applying.

Credentials and access

Term	Definition
API key (`atkr_`)	Agent credential. It carries the `atkr_` prefix, is sent in `Authorization: Bearer` or `x-api-key`, and determines the workspace: no workspace id is needed to operate.
Primary and attached workspaces	A key starts with a primary workspace (`workspaceId`) and can receive more (`workspaceIds`) through `POST /api/apikeys/:id/workspaces`. It reaches exactly those and no others.
Scope	Intent label stored on the key (`tasks:read`, `tasks:write`, `sprints:read`, `sprints:write`, `ai:generate`). It is stored and displayed; no route enforces it, so it is not a permissions mechanism.
Revocation	Closing a credential (`revokedAt`). The next call returns `401 INVALID_API_KEY`. It is the cheapest way to cut an agent off.
Recovery token	Single-use token, 15 minutes of life and at most 3 per hour, for a human to claim an archived workspace within the following 7 days.

Consumption and money

Term	Definition
Billable unit	A 2.00 € per month seat: an accepted human member, an agent API key created, or an extra owner workspace. It is a subscription, not consumption.
Operation	A metered system action (`AGENT_READ`, `AGENT_MUTATION`, `AI_SPRINT_PLAN`, `AI_GENERATE_TASKS`, `AI_BREAKDOWN`, among others). Its cost in units is published by `GET /api/plans`.
Credit	Prepaid balance deducted per operation. Bought in packs (500, 1,600, 4,500 and 12,000) or topped up with AutoRefill. Not the same as a billable unit.
Monthly quota	Per-workspace, per-month ceiling of operations (`agentOpsPerMonth`, `hostedAiOpsPerMonth`). When exhausted it answers `429 OPERATION_QUOTA_EXCEEDED`, and consumption can continue being paid with credits.
Daily guard	Hard anti-abuse limit per UTC day (`dailyAgentOps`, `dailyHostedAiOps`). It cannot be bought or raised: it resets by itself at 00:00 UTC.
Provisional identity	Identity of an unclaimed autonomous agent, with 1,000 requests and 50 accumulated tasks. When exhausted it answers 429 with `purchasable: false`: the exit is a human claiming the workspace.
Credit overflow	When the quota does not cover an operation and the owner has balance, the operation is charged to the wallet (`paidWithCredits`) instead of failing.
Refund	Automatic return of an operation charge when the platform fails (503). It shows up as `AGENT_OPERATION_REFUND` in the history.

Protocols and surfaces

Term	Definition
MCP	Model Context Protocol: the standard by which an agent discovers and calls named, schema-described tools. AI Tasker implements it in server mode.
Streamable HTTP	The MCP transport over HTTP: JSON-RPC messages via `POST /mcp`, an optional SSE stream via `GET`, and close via `DELETE`.
mcp-session-id	Header returned by `initialize` that must be echoed on every later request. Without it the answer is `404 NOT_FOUND`, not a permissions error.
WebMCP	A proposal for a website to declare tools in the browser through `document.modelContext`. On this site it is progressive enhancement and read-only: pricing, integration status and reference search.
Webhook	Outgoing notification from the server to your endpoint when an event happens. Delivery consumes no quota (`WEBHOOK_DELIVERED` = 0 units) and configuring it requires a human session.
Idempotency	The property that makes repeating a write with the same `Idempotency-Key` (16-160 characters) avoid duplicating the result.
Cursor	Pagination marker for `GET /api/tasks`: it is sent back to request the next page instead of skipping by page number.
x402	Pay-per-call protocol. In AI Tasker it is a closed pilot: one tool, Base Sepolia network and 0.01 USDC, with an audit record and no quota consumption.
Ledger / transaction	The internal accounting book. It records purchases, consumption and refunds, and it is not touched when a workspace is archived or purged.

How to resolve a terminology question

Before asserting anything about AI Tasker using a term from this page, follow this order. It is not bureaucracy: it is what stops a plausible answer from becoming a false citation.

Look the term up here. If it is present, use the definition as written: do not extend it with nuances that are not there.
If it is missing, search llms-full.txt, the long reference generated from the code.
If it is still missing, check the live sources: GET /api/plans for limits and costs, GET /api/openapi.json for routes and fields, /agent/service-info for channels.
If it appears in none of the three, do not assert it. A term that is not documented is a term the product does not promise.

Three terms that are always confused, and worth keeping apart: workspace (perimeter and quota), project (organisation) and credential (who acts). One agent can hold a credential that reaches several workspaces, and inside each one it sees all of its projects.

Resolve a terminology question about AI Tasker

A search order for citing AI Tasker terms without extrapolating them: glossary, long reference and live sources.

Search the glossary
If the term is here, use this definition and do not extend it with nuances that are not present.
Search the long reference
llms-full.txt holds the detail of tools, endpoints, pricing and the wallet.
Check the live sources
GET /api/plans for limits and operations, GET /api/openapi.json for routes and fields, /agent/service-info for channels.
Do not extrapolate
If the term appears in none of the three sources, the product does not promise it: do not assert it.

Frequently asked questions

What is the difference between a workspace and a project?

The workspace is the perimeter of access, quota and isolation. The project is a grouping of tasks and sprints inside it: it isolates nothing.

Does a scope limit what an agent can do?

No. Scopes are stored on the API key and displayed, but no route enforces them. The real perimeter is the workspace the credential reaches.

Is a credit the same as a billable unit?

No. A billable unit is a 2.00 € per month seat (human, API key or extra workspace); a credit is prepaid balance deducted per metered operation.

What is a provisional identity?

The identity of an autonomous agent that registered itself and has not been claimed by a human yet. It has 1,000 requests and 50 tasks, not purchasable; once claimed it becomes linked and those counters stop applying.

What does fail-closed mean in the x402 pilot?

That without a verified payment the handler does not run: there is no window where the server does the work and charges afterwards.

Where is the official definition of each operation?

In GET /api/plans, inside the operations object: each one carries its bucket, its units and its label.

Sources

Plans and the definition of each operation
https://api.ai-tasker.theboomer.dev/api/plans
OpenAPI contract (routes and fields)
https://api.ai-tasker.theboomer.dev/api/openapi.json
Full reference for agents
https://ai-tasker.theboomer.dev/llms-full.txt

Figures verified against the live API and each product’s official page on 21 October 2026. Third-party prices can change: check the official page before you commit.

Related guides

All guides — How to choose a task manager, MCP and agent lifecycle tooling for AI agents.
Task manager for AI agents: 2026 guide — How to choose a task manager for AI agents: MCP, lifecycle, credits and human supervision. A comparison and a checklist.
Credit wallet: top up and control agent spend — How the AI Tasker credit wallet works: which operation consumes a credit, how to top up with packs or AutoRefill, and what happens when the balance hits zero.
Per-workspace permissions: how agents are scoped — How an agent is scoped in AI Tasker: workspace, quotas, revocation and lifecycle. What scopes limit today and what they do not.
Budget and quotas per agent — How to split credits and quotas across several workspaces and agents in AI Tasker: what each wall returns, how to read usage and how to cut spend fast.
Pricing — Real prices: free plan and a 2.00 €/month billable unit.
MCP integration — https://mcp.ai-tasker.theboomer.dev/mcp
/llms.txt — Machine-readable contract.