# Powerloom > Powerloom is a control plane for AI agent fleets. IT teams declare agents, the tools they can reach, and the policies that govern them — in YAML manifests, applied through a `weave plan` / `weave apply` workflow that mirrors Terraform. Every state-changing action lands in an append-only, SHA-256 hash-chained audit log. The model is deliberately Active Directory-shaped: organizational units, security groups, role bindings with deny-override. ## Docs for operators - [Getting started](https://powerloom.org/docs/getting-started/raw): Sign up, model your first organizational unit, deploy your first agent, and bind a role. Twenty minutes from a fresh workspace to a governed agent in production. - [OU model & RBAC](https://powerloom.org/docs/rbac/raw): Organizational units, security groups, role bindings, and deny-override. The Active Directory model, applied to agents and the tools they reach. - [Agent deployment](https://powerloom.org/docs/agent-deployment/raw): Declare an agent in YAML, attach skills and MCP servers, set a scope, apply. Reconciler drives the rollout; audit log captures the decision. - [Auth & audit](https://powerloom.org/docs/auth-and-audit/raw): OIDC SSO, JWT issuance, personal access tokens, and the SHA-256 hash-chained audit log. Every state-changing action lands in a tamper-evident record. - [MCP deployment](https://powerloom.org/docs/mcp-deployment/raw): Pre-built templates for Postgres, Slack, GitHub, Files, and more. Deploy via manifest, scope by OU, audit every tool call. - [Workflows](https://powerloom.org/docs/workflows/raw): DAG-shaped multi-agent handoffs declared in YAML. Seven node kinds: trigger, agent, condition, approval, subflow, output, transform. ## Docs for developers - [CLI reference](https://powerloom.org/docs/cli/raw): weave: the command-line surface for plan, apply, diff, destroy, get, describe. Every flag, every exit code. Installs with `pip install loomcli`. - [loomcli (the package)](https://powerloom.org/docs/loomcli/raw): The Python package that ships the weave binary + the bundled v2 JSON Schema. Install paths, version pinning, build-from-source, and the schema-source-of-truth role. - [Manifest schema](https://powerloom.org/docs/manifest-schema/raw): YAML manifest shape for every Kind: OU, Group, RoleBinding, Agent, Skill, MCPDeployment, MCPServerRegistration, Credential. Fields, defaults, validation. - [API reference](https://powerloom.org/docs/api-reference/raw): REST endpoints for every resource. JWT and PAT auth, OIDC SSO, pagination, error shape. Full OpenAPI spec at /api/openapi.json. - [Alfred (meta-agent)](https://powerloom.org/docs/meta-agent/raw): The coordinator that drafts manifest diffs from natural language. How to invoke, what it reads from memory, and how to bound its scope. - [Tracker tools](https://powerloom.org/docs/tracker-tools/raw): MCP tool definitions for the project tracker — list_open_threads, claim_thread, comment_on_thread, close_thread, search_similar_threads, get_thread_detail. ## Optional - [Marketing landing](https://powerloom.org/): What Powerloom is, in marketing voice - [Why Powerloom](https://powerloom.org/why-powerloom): Long-form positioning brief - [Security & trust](https://powerloom.org/security): Trust posture + roadmap - [Full docs corpus](https://powerloom.org/llms-full.txt): Every doc concatenated for one-shot ingestion