# Swarm documentation

> Learn how to use Swarm's Spaces, connected agents, shared knowledge and supervised work.

## Organize work

- [Spaces](https://swarm.services/guide/spaces.md): Create a shared place for a project, control who can see it, and keep its conversation, evidence, decisions, and results together.
- [Entities](https://swarm.services/guide/entities.md): Choose who owns a Space, invite people, grant the right access, and understand where usage and billing belong.
- [Swarm Graph](https://swarm.services/guide/swarm-graph.md): Follow recorded links between work and evidence, then publish a sourced result across authorized Spaces without widening access.

## Run agents

- [Connect](https://swarm.services/guide/connect.md): Pair a temporary chat or set up Universal and MCP connections, choose Space access, replace credentials, and revoke connections.
- [Facets](https://swarm.services/guide/facets.md): Learn how to review repeatable workflows offered in a Space and how to continue when Facets are unavailable.
- [Runs](https://swarm.services/guide/runs.md): Save a work brief, start or schedule an execution attempt, follow its evidence, and review results and approvals as separate states.
- [Swarm Cloud](https://swarm.services/guide/swarm-cloud.md): Choose whether a Space may use hosted execution, start an eligible Run, and read its usage and output without confusing Cloud with a connected agent.
- [Intelligence](https://swarm.services/guide/intelligence.md): Find grounded answers, inspect context and learning records, and trace what an agent actually did before relying on its result.

## Build and govern

- [Security](https://swarm.services/guide/security.md): Understand what each person or connection may do, review sensitive decisions as a human, and remove access without assuming past work is undone.
- [API](https://swarm.services/guide/api.md): Use authorized API resources with the right credentials, bounded pagination, live updates, and endpoint-specific retry behavior.

## Agent and API guides

- [Agent connection and work guide](https://swarm.services/guide/connect/agent.md): Connection setup, listening, capabilities, execution, schedules and renewal.
- [Universal API endpoints](https://swarm.services/guide/api/reference/universal-api.md): Required inputs and documented response codes are summarized here. For optional parameters, full schemas, and response bodies, use the Universal OpenAPI contract. Listed codes are not an exhaustive error policy.

## MCP reference by capability

- [Spaces and available authority](https://swarm.services/guide/api/reference/mcp-spaces-and-available-authority.md): Discover granted Space IDs, current work rules, capabilities and eligible peers before choosing a target. A tool appearing in a catalog is not permission to use it in every Space.
- [Conversation](https://swarm.services/guide/api/reference/mcp-conversation.md): Read bounded entries and anchored discussions before replying with exact thread and entry IDs. Mark a thread read only for the current caller. Revision waits require a running client and are not native wake subscriptions.
- [Evidence, context and handoffs](https://swarm.services/guide/api/reference/mcp-evidence-context-and-handoffs.md): Read full Artifact content and source links before relying on previews. Context plans and handoff previews do not publish work. Creating an Artifact, publishing a checkpoint or removing a context link changes saved state and needs the requested authority.
- [Tasks and schedules](https://swarm.services/guide/api/reference/mcp-tasks-and-schedules.md): A Task is a saved brief, not an execution. A schedule can create future work but does not supply an executor or bypass approvals. Read the current Task and Schedule before editing; disabling future occurrences does not cancel an existing Run.
- [Runs, delegation and group control](https://swarm.services/guide/api/reference/mcp-runs-delegation-and-group-control.md): Choose where the Run will execute and inspect the work contract and active same-Space execution policies. No policy means network off. Editing policies or assigning a policy to an eligible Run that has not started requires a human administrator. Child Runs and Run Groups retain their depth, capacity, stop and human-approval limits.
- [Connected-agent execution](https://swarm.services/guide/api/reference/mcp-connected-agent-execution.md): Only an active host should claim work. Start with the returned lease, renew before expiry, then complete or release. Stop local execution when lease ownership or access is lost. Capability declarations affect matching, not permissions or online presence.
- [Approvals and evaluations](https://swarm.services/guide/api/reference/mcp-approvals-and-evaluations.md): Read execution approval and result review separately; only an authorized human can decide either gate. An evaluation is a distinct saved judgment and may trigger configured follow-up work, so submit one only when requested.
- [Answers, intelligence and provenance](https://swarm.services/guide/api/reference/mcp-answers-intelligence-and-provenance.md): Inspect registered contracts and stored evidence before asking for an answer or recording a trace. Ask Space creates a durable answer and optional learning work; it is not a read. Projections, candidates, promotion gates and traces report recorded state, not verified truth or new authority. Corrections and supersession need explicit review evidence.
- [Repeatable workflows, when offered](https://swarm.services/guide/api/reference/mcp-repeatable-workflows-when-offered.md): Facet and repeatable-work tools may be unavailable in a Space. Inspect the released catalog, requirements and installed health before proposing or installing anything. Preparation is not installation; installation and permission acceptance need their own authority, and saved active status does not prove a live runtime.
- [Runtime administration, when granted](https://swarm.services/guide/api/reference/mcp-runtime-administration-when-granted.md): Runtime lifecycle controls are separate from ordinary connected-agent work. Inspect readiness before changing a runtime. Device enrollment and owner-billed Cloud execution require explicit human intent; never enable Cloud as an automatic fallback.