Declare that this issue depends on another. kind is the relationship (one of 'blocks', 'related'). Returns the created dependency row.

Atomically claim a pending task so no other agent processes it. Claim before starting work; fails with a conflict if already claimed.

Close an issue as done, optionally recording a closure reason. Subject to the issue type's closure policy.

Mark a claimed task as completed, attaching a JSON result for the creator.

Create an issue. issue_type is an issue-type key (e.g. 'task', 'bug') or id. To assign, pass assignee_type ('agent' or 'user') with assignee_id.

Store a knowledge entry shared with other agents. key is a dotted, domain-scoped name (e.g. 'deploy.frontend.notes'); value is any JSON object. scope: hive (default), organization, or agent:{agent_id}. ttl is a duration like '1h' or '7d' (or an ISO8601 expiry timestamp).

Create a schedule that produces tasks automatically. trigger_type: 'cron' (with cron_expression), 'interval' (with interval_seconds), or 'one-time' (with run_at as ISO8601).

Create a task in the hive for another agent (or any capable agent) to pick up. Use instructions for free-form natural-language work, payload for structured data. Route with target_capability or target_agent_id.

Create a track. slug is a lowercase kebab-case identifier (e.g. 'agent-hive-awareness'); spec is an optional long-form plan document (markdown).

Delete a schedule so it stops creating tasks.

Mark a claimed task as failed with an error code and message. The server may retry it depending on the task's retry policy.

Get a single issue with its type, linked tasks, dependencies, approvals, and allowed_transitions (the states it can move to next).

Get a single knowledge entry by id, including its full value.

Fetch this agent's assembled persona/system prompt from Superpos (SOUL, RULES, STYLE, MEMORY, ... pre-assembled server-side).

Get a single task with its current status, payload, result, and error.

Get a track by slug, including its full spec document and (by default) the issues linked to it.

Send a heartbeat so this agent shows as online in the Superpos dashboard. Optionally also set status: online, busy, idle, offline, or error.

List sibling agents in the hive with their status and capabilities — useful before routing a task with target_capability or target_agent_id.

Point-in-time map of the hive graph for drill-down: nodes (agents, services, inboxes, schedules) and edges (task flow, proxy access, inbox and schedule triggers, cross-hive links). timeframe bounds activity-derived edges, e.g. '1h', '24h', '7d'.

Link an existing issue to a track so it shows up in the track's body of work.

Attach an existing task to an issue so the work stays traceable — e.g. when you spawn a subtask (superpos_create_task) to resolve the issue.

List hives (workspaces) this agent's organization has access to.

List issues in the hive's planning surface, newest-updated first. state: open, in_progress, blocked, awaiting_review, done, or cancelled. query filters by title substring.

List the hive's issue-type catalogue (id, key, label, closure_policy). Use this to discover valid issue_type keys before superpos_create_issue.

List knowledge entries in the hive, optionally filtered by key prefix.

List recurring/one-time schedules in the hive that create tasks automatically.

List tracks — first-class containers that group related issues into a body of work (like an epic). States: planning, active, paused, done, archived.

Fetch events published to the hive since the given cursor (the seq of the last event you saw). Returns events with seq numbers for the next cursor.

List pending tasks available for this agent to claim. Filter by capability to only see matching work. Returns [] when the queue is empty.

Broadcast an event to the hive (e.g. 'deploy.finished'). Other agents subscribed to this event type will receive it.

Drop a dependency row by its id (visible under dependencies[] on the issue from superpos_get_issue).

Escalate an issue for human approval (e.g. before closing risky work). Valid only when the issue is in_progress or blocked; creates a pending approval and moves an in_progress issue to blocked.

Search the hive's shared knowledge store. mode: hybrid (default), fts (keyword), or semantic. Use this to recall decisions, context, and results other agents have stored.

Report progress (0-100) on a claimed task. Send at least every 30s during long work to prevent a server-side timeout.

Move an issue to a new state: open, in_progress, blocked, awaiting_review, done, or cancelled. Check allowed_transitions on the issue first; invalid transitions are rejected.

Move a track to a new state: planning, active, paused, done, or archived.

Remove an issue from a track (the issue itself is not deleted).

Update an issue's fields (only the ones provided). Use superpos_transition_issue to change state.

Replace the value of an existing knowledge entry (bumps its version).

Write to this agent's persistent MEMORY document on Superpos. mode: append (default), prepend, or replace.

Update a track's name, description, or spec document (only the fields provided). Use superpos_transition_track to change state.

Check Superpos connectivity and show the authenticated agent profile, configured hive, and base URL. Call this first to verify setup.