Read-only. Search the task store by free-text QUERY using available adapter signals plus Core keyword retrieval, fused via Reciprocal Rank Fusion. Adapters with embeddings add dense+sparse hybrid retrieval. Returns best-matching items with provenance. Use this for "what do I have about X". To find items like a KNOWN item instead, use similar. Read-only: never writes.

Read-only. Fetch one item (task/note) by its project id + task id, including the full markdown body, tags, and due date. Use after find/list_projects give you the ids. Returns the single item object, or an error if not found.

Read-only. List the projects / folders in the store, each with its id and name. Call this first to discover the projectId you need for get_task, create_task, or update_task. Returns an array of { id, name }.

WRITE. Creates a new item (task/note) in the store — the agent→human write side of the two-way bus. Use to hand the human a note, action, or reminder. Side effect: a new item appears in their task app immediately. Returns the created item including its new id. To change an existing item instead, use update_task.

WRITE. Patches an existing item (partial update) — only the fields you provide change; omitted fields are left untouched. Side effect: the item is modified in the human's task app. Requires the item's projectId + taskId (get them from find or list_projects). Returns the updated item. To create a new item instead, use create_task.

WRITE. Adds or updates portable execution intent inside the task body: desired outcome, why it matters, completion conditions, authority, constraints, and approval requirement. Preserves the human-authored body and works through every ATS adapter.

WRITE. Sets portable lifecycle state and validity windows. Archived, expired, future, and superseded tasks are excluded by context assembly.

WRITE. Creates a committed execution item from exploratory material. The source body is not copied; the new item receives explicit outcome/completion metadata and an evidence link back to the source.

Read-only. Returns the item role and its explicit parent relationship. Roles are exploration, goal, project, task, or unspecified.

WRITE. Assigns an exploration/goal/project/task role and optionally replaces or clears the single explicit parent relationship.

Read-only. Deterministically checks whether a task still supports its parent objective and reports invalid role ordering, missing intent, cycles, invalid lifecycle state, and active explicit conflicts.

Read-only. Returns the portable trust, action, resource, denial, approval, and approver policy for one task. Unconfigured tasks default to untrusted content and no granted access.

WRITE. Sets a portable task security policy. This policy is a decision point for cooperating clients; it does not sandbox unrelated external tools.

WRITE/AUDIT. Evaluates one task-scoped access request and appends an allow or deny record. Requires an action, resource, and reason. Denial wins over allowance; untrusted high-risk actions and configured boundaries require approval. Fails closed if auditing fails.

WRITE. Adds a typed relationship from one task to another. Types express dependencies, evidence, decisions, outputs, supersession, support, or a general relation.

WRITE. Removes one exact typed relationship from a source task. Returns removed=false when the relationship was already absent.

Read-only. Traverses typed incoming and outgoing task relationships around one task. Returns nodes, edges, lifecycle validity, and unresolved references.

Read-only. Builds execution context for a task. Typed relationships come first, retrieval adds candidates, invalid lifecycle items are excluded, and every included item carries provenance.

WRITE. Appends an auditable agent action or outcome to the local ATS JSONL ledger, including sources, approvals, output, and whether the task advanced.

Read-only. Lists action-ledger records, optionally filtered by task, agent, action, or advancement.

WRITE/LOCAL. Creates or replaces the local corpus-diff event checkpoint. This only observes task state; it never launches an agent or performs an external action.

WRITE/LOCAL. Diffs the current task corpus, durably stages deterministic event envelopes, then advances the checkpoint. Returns newly observed events plus the unacknowledged pending spool. Events are observation-only and must be acted on separately under normal ATS intent and security checks.

Read-only. Lists durable task-event envelopes that have not been acknowledged by a consumer. Task bodies are not stored in the spool.

WRITE/LOCAL. Explicitly acknowledges one or more durable task-event ids and removes them from the pending spool. Unknown ids are reported without failing known acknowledgements.

Read-only. Find items semantically similar to a KNOWN item, given its id (not a text query — for text search use find). Useful for "show me related notes to this one" or dedupe. Requires an embedder-backed adapter (e.g. TickTick); returns a clear error on adapters without embeddings. Returns an array of items ranked by similarity.

Read-only. Build a paste-ready deep link (a URL string) to an item in its native app, so you can hand the human a clickable reference. Does not open anything or write — pure id→URL construction. Returns the URL as a string.