# Meerkat

> A modular, high-performance agent harness built in Rust

## Docs

- [MCP](https://docs.rkat.ai/api/mcp.md): Use Meerkat as an MCP client to call external tools, or expose Meerkat as MCP tools for other clients.
- [REST API](https://docs.rkat.ai/api/rest.md): HTTP server for running and managing agent sessions over REST endpoints.
- [JSON-RPC stdio server](https://docs.rkat.ai/api/rpc.md): Stateful JSON-RPC 2.0 interface for stdio, TCP, and optional live audio WebSocket hosting.
- [Meerkat Dogma](https://docs.rkat.ai/architecture/meerkat-dogma.md): Primary runtime architecture doctrine for semantic authority, seams, projections, and surfaces.
- [Meerkat Dogma Commentary](https://docs.rkat.ai/architecture/meerkat-dogma-commentary.md): Companion commentary with examples, repair shapes, and review questions for Meerkat dogma.
- [CLI commands](https://docs.rkat.ai/cli/commands.md): Command reference for the task-first, Unix-native rkat CLI.
- [CLI configuration](https://docs.rkat.ai/cli/configuration.md): Realm selection, storage behavior, environment variables, MCP config, and exit codes for rkat.
- [Auth And Bindings](https://docs.rkat.ai/concepts/auth-and-bindings.md): How Meerkat separates provider selection from credential resolution and binding identity.
- [Comms](https://docs.rkat.ai/concepts/comms.md): How long-lived agents exchange messages and external events in Meerkat.
- [Composability](https://docs.rkat.ai/concepts/composability.md): Why Meerkat is both a composable library and a practical product layer.
- [Configuration](https://docs.rkat.ai/concepts/configuration.md): Realm-scoped config, generation CAS, and surface-consistent runtime settings.
- [Memory And Compaction](https://docs.rkat.ai/concepts/memory-and-compaction.md): The long-horizon conversation model: context compaction plus semantic recall.
- [Mobs](https://docs.rkat.ai/concepts/mobs.md): Meerkat's multi-agent model: durable teams, stable member identity, profile-based roles, wiring, flows, and host-vs-agent surfaces.
- [Concepts Overview](https://docs.rkat.ai/concepts/overview.md): The conceptual map of Meerkat: composability, identity, lifecycle, capabilities, and orchestration.
- [Providers](https://docs.rkat.ai/concepts/providers.md): Provider-agnostic: same tools, same sessions across Anthropic, OpenAI, Gemini, and configured self-hosted models.
- [Realms](https://docs.rkat.ai/concepts/realms.md): Realm-first identity and storage model across CLI, RPC, REST, MCP, and SDK surfaces.
- [Scheduling](https://docs.rkat.ai/concepts/scheduling.md): Durable automated work over time: triggers, targets, occurrences, and delivery policies.
- [Sessions](https://docs.rkat.ai/concepts/sessions.md): Realm-scoped session lifecycle with deterministic concurrency and cross-surface consistency.
- [Tools](https://docs.rkat.ai/concepts/tools.md): Trait-based tool dispatch, MCP server integration, and built-in tool categories.
- [WorkGraph](https://docs.rkat.ai/concepts/workgraph.md): Realm-scoped durable commitments, dependencies, readiness, claims, and evidence.
- [Comms & external events](https://docs.rkat.ai/examples/comms.md): Keep-alive sessions, typed comms commands, peer discovery, and durable external event ingress.
- [Examples gallery](https://docs.rkat.ai/examples/gallery.md): Functional apps and production patterns built with Meerkat. From coding swarms to browser war games.
- [Hooks](https://docs.rkat.ai/examples/hooks.md): Lifecycle hooks for audit and guardrails at 8 hook points.
- [Memory & compaction](https://docs.rkat.ai/examples/memory.md): Semantic memory indexing, context compaction, and budget controls.
- [Mobpack](https://docs.rkat.ai/examples/mobpack.md): Portable multi-agent deployment with signed archives and trust policies.
- [Mobs](https://docs.rkat.ai/examples/mobs.md): Create mobs, spawn members, wire peers, send work, and run flows through the current host control plane.
- [Sessions & streaming](https://docs.rkat.ai/examples/sessions.md): Session lifecycle, multi-turn conversations, event streaming, model selection, and config management.
- [Skills](https://docs.rkat.ai/examples/skills.md): Load, discover, inject, and inspect domain-specific knowledge packs.
- [Structured output](https://docs.rkat.ai/examples/structured-output.md): JSON schema extraction with validation, retries, and provider compatibility.
- [Tools & MCP](https://docs.rkat.ai/examples/tools.md): Custom tools, built-in tools, MCP server management, live MCP controls, and tool scoping.
- [WASM runtime](https://docs.rkat.ai/examples/wasm.md): Browser deployment with the meerkat-web-runtime embedded surface.
- [WorkGraph Example](https://docs.rkat.ai/examples/workgraph.md): A realistic multi-step workflow using dependencies, claims, evidence, and readiness.
- [Auth](https://docs.rkat.ai/guides/auth.md): Realm-scoped credential resolution: from the env-var happy path to OAuth, cloud IAM, and per-member overrides
- [CD and distribution](https://docs.rkat.ai/guides/cd-and-distribution.md): How Meerkat release artifacts are validated, built, and published across Rust, binaries, Python, and TypeScript.
- [Inter-agent communication](https://docs.rkat.ai/guides/comms.md): Built-in messaging system for coordination between multiple agent instances
- [Hooks](https://docs.rkat.ai/guides/hooks.md): Typed extension points within the agent loop for observation and guardrails
- [Image generation](https://docs.rkat.ai/guides/image-generation.md): Generate or edit assistant-owned images through the session-scoped generate_image tool.
- [Memory and compaction](https://docs.rkat.ai/guides/memory.md): Automatic context compaction and semantic memory for long-running conversations
- [Mobpack and Web Deployment](https://docs.rkat.ai/guides/mobpack.md): Portable multi-agent deployment with signed .mobpack artifacts and browser-target web bundles.
- [Mobs](https://docs.rkat.ai/guides/mobs.md): Build and operate Meerkat multi-agent teams: delegation, members, wiring, flows, work lanes, profiles, and mobpacks.
- [Realm guide](https://docs.rkat.ai/guides/realms.md): How rkat chooses a realm, where realm state lives, and how to share or isolate state deliberately.
- [Live Channels](https://docs.rkat.ai/guides/realtime.md): Open live audio/text channels on sessions using realtime-capable models and the live/* RPC surface.
- [Scheduling](https://docs.rkat.ai/guides/scheduling.md): Automate durable session and mob work with once, interval, and calendar schedules.
- [Self-hosting models](https://docs.rkat.ai/guides/self-hosting-models.md): Register OpenAI-compatible local or private model servers, including the Gemma 4 aliases Meerkat supports.
- [Skills](https://docs.rkat.ai/guides/skills.md): Domain-specific knowledge injection via markdown documents with YAML frontmatter
- [Structured output](https://docs.rkat.ai/guides/structured-output.md): Force agents to produce validated JSON conforming to a user-provided schema
- [WorkGraph](https://docs.rkat.ai/guides/workgraph.md): Use WorkGraph tools to create, link, claim, evidence, close, and observe durable work.
- [Introduction](https://docs.rkat.ai/introduction.md): Meerkat is a library-first agent engine built in Rust. Provider-agnostic, modular, fast.
- [Quickstart](https://docs.rkat.ai/quickstart.md): Install Meerkat and run your first agent. Single binary, any provider.
- [API reference](https://docs.rkat.ai/reference/api-reference.md): Index of public types, traits, and error codes in Meerkat.
- [Architecture](https://docs.rkat.ai/reference/architecture.md): A high-level map of Meerkat's runtime, surfaces, machines, storage, providers, and orchestration model.
- [Build And CI](https://docs.rkat.ai/reference/build-and-ci.md): The Make, Cargo, BuildBuddy, and release validation surfaces used by Meerkat contributors.
- [Built-in tools reference](https://docs.rkat.ai/reference/builtin-tools.md): Comprehensive reference for all built-in tools available in Meerkat, including parameters, types, and defaults.
- [Capability matrix](https://docs.rkat.ai/reference/capability-matrix.md): Build profiles, feature gates, error codes, and capability behavior across Meerkat configurations.
- [Comms Reference](https://docs.rkat.ai/reference/comms-reference.md): Low-level comms reference: agent-facing tools, host-side comms/send surfaces, peer discovery, and ingress distinctions.
- [Design Philosophy](https://docs.rkat.ai/reference/design-philosophy.md): The engineering principles behind Meerkat's runtime, API boundaries, Rust implementation style, and validation model.
- [Hooks Reference](https://docs.rkat.ai/reference/hooks-reference.md): Low-level hook surface reference: hook points, capabilities, execution modes, and payload fields.
- [Machine Authority](https://docs.rkat.ai/reference/machine-authority.md): The canonical machine catalog, generated artifacts, and validation gates that keep runtime semantics aligned with code.
- [Mob Architecture](https://docs.rkat.ai/reference/mob-architecture.md): How Meerkat models multi-agent orchestration, member identity, runtime bindings, wiring, flows, and mobpacks.
- [Runtime Architecture](https://docs.rkat.ai/reference/runtime-architecture.md): How Meerkat routes session work through runtime-backed surfaces, session services, and the agent factory.
- [Session contracts](https://docs.rkat.ai/reference/session-contracts.md): Canonical session, runtime, keep-alive, and commit-boundary contracts.
- [Skills Governance Runbook](https://docs.rkat.ai/reference/skills-governance-runbook.md): Operational runbook for source UUID lifecycle, lineage migrations, and remap governance.
- [Skills Reference](https://docs.rkat.ai/reference/skills-reference.md): Low-level skill runtime reference: source transports, identity, injection paths, and discovery tools.
- [Structured Output Reference](https://docs.rkat.ai/reference/structured-output-reference.md): Low-level structured-output reference: schema types, compatibility modes, warnings, and result fields.
- [WorkGraph Reference](https://docs.rkat.ai/reference/workgraph-reference.md): WorkGraph item fields, statuses, edges, evidence, tools, and read-only host surfaces.
- [Rust SDK advanced](https://docs.rkat.ai/rust/advanced.md): Expert-level direct agent construction, provider configuration, budgets, hooks, and runtime delivery internals.
- [Rust SDK overview](https://docs.rkat.ai/rust/overview.md): Embed Meerkat as a Rust library: the full engine with no subprocess overhead.
- [Tools and stores](https://docs.rkat.ai/rust/tools-and-stores.md): Rust SDK reference for the tool system, session stores, and MCP integration.
- [Python SDK](https://docs.rkat.ai/sdks/python/overview.md): Get started with the Meerkat Python SDK: sessions, streaming, schedules, mobs, and typed runtime contracts.
- [Python SDK reference](https://docs.rkat.ai/sdks/python/reference.md): Public types and behavior for the Meerkat Python SDK.
- [TypeScript SDK](https://docs.rkat.ai/sdks/typescript/overview.md): Getting started with the Meerkat TypeScript SDK: sessions, turns, streaming events, and capabilities.
- [TypeScript SDK reference](https://docs.rkat.ai/sdks/typescript/reference.md): Types, events, error classes, and version compatibility for the @rkat/sdk TypeScript SDK.

## OpenAPI Specs

- [openapi](https://docs.rkat.ai/api-reference/openapi.json)