# Kheish daemon

> Persistent agent runtime and control plane for long-lived, tool-using AI workflows.

## Docs

- [Agents and sidechains](https://docs.kheish.ai/concepts/agents-and-sidechains.md): Agent hierarchy, sidechains, mailboxes, and profile-constrained execution.
- [Approvals and structured questions](https://docs.kheish.ai/concepts/approvals-and-questions.md): Human-in-the-loop suspension points for sensitive actions and explicit user input.
- [Assets and multimodal input](https://docs.kheish.ai/concepts/assets-and-multimodal-input.md): Daemon-owned file storage, supported media types, and the input shapes used for text, documents, images, and retained audio.
- [Capture and observations](https://docs.kheish.ai/concepts/capture-and-observations.md): How Kheish models screenshots, webcam frames, microphone segments, and other captured inputs as durable daemon-owned observations.
- [Channels and public conversations](https://docs.kheish.ai/concepts/channels-and-public-conversations.md): Durable shared channels, canonical root threads, reactions, and stimulus-driven public turns over session-backed agents.
- [Connectors and reply targets](https://docs.kheish.ai/concepts/connectors-and-reply-targets.md): Daemon-managed transport resources, session reply defaults, and the run snapshots that make output routing durable.
- [Personas](https://docs.kheish.ai/concepts/personas.md): Mutable persona records, immutable session bindings, sidechain inheritance, and the prompt semantics that make personas durable.
- [Playbooks and flows](https://docs.kheish.ai/concepts/playbooks-and-flows.md): Declarative Playbook versions and Flow projections over normal Kheish sessions, runs, agents, and approvals.
- [Projects and project tasks](https://docs.kheish.ai/concepts/projects-and-project-tasks.md): Daemon-owned project coordination, durable project members, linked channels, and project-task execution.
- [Recovered run memory](https://docs.kheish.ai/concepts/recovered-run-memory.md): Daemon-owned episodic memory derived from terminal runs, with bounded prompt recovery, durable indexing, and scope-aware search.
- [Semantic learning and procedural skills](https://docs.kheish.ai/concepts/semantic-learning-and-procedural-skills.md): Durable daemon-owned learnings, automatic capture and review, bounded prompt retrieval, and promotion of reviewed procedures into reusable skills.
- [Sessions and runs](https://docs.kheish.ai/concepts/sessions-and-runs.md): Durable sessions, queued runs, and the separation between conversation state and concrete execution.
- [State and recovery](https://docs.kheish.ai/concepts/state-and-recovery.md): Journals, checkpoints, compaction, and the restore model that makes the daemon durable.
- [Tasks and schedules](https://docs.kheish.ai/concepts/tasks-and-schedules.md): Structured work tracking, background shell tasks, and scheduled execution inside the daemon.
- [Development](https://docs.kheish.ai/contributing/development.md): Repository setup, testing strategy, and contributor workflow for the Kheish daemon project.
- [Kheish daemon](https://docs.kheish.ai/index.md)
- [Capture One Observation](https://docs.kheish.ai/integrations/capture-quickstart.md): Use the host-local capture runtime to provision one observation source and upload one fixture observation into the daemon.
- [Connectors](https://docs.kheish.ai/integrations/connectors.md): Daemon-managed ingress connectors for HTTP, Slack, Telegram, and external sidecars.
- [Multiple input connectors in one session](https://docs.kheish.ai/integrations/multi-input-sessions.md): How to route several connectors into the same durable session, why to do it, and what behavior to expect.
- [Output routing](https://docs.kheish.ai/integrations/output-routing.md): Daemon-local output persistence, reply targets, generic output plugins, and queued delivery.
- [Send One Webhook Event Into Kheish](https://docs.kheish.ai/integrations/webhook-quickstart.md): Create one webhook connector, bind it to a fixed session, and verify that one external event becomes one daemon run.
- [Architecture overview](https://docs.kheish.ai/introduction/architecture-overview.md): The main subsystems that make up Kheish and how requests flow through them.
- [Overview](https://docs.kheish.ai/introduction/overview.md): What Kheish is, what it is for, and how the daemon-first model changes agent operations.
- [Use Kheish From Python](https://docs.kheish.ai/introduction/python-sdk-quickstart.md): Connect one Python program to a running Kheish daemon and complete one real response run.
- [Quickstart](https://docs.kheish.ai/introduction/quickstart.md): Get one working Kheish daemon, one session, and one completed run, either from the local binary or through Docker.
- [Start Here](https://docs.kheish.ai/introduction/start-here.md): Understand what runs where in the Kheish ecosystem and pick the shortest path to your first real success.
- [Web SDK quickstart](https://docs.kheish.ai/introduction/web-sdk-quickstart.md): Browser integration patterns for the Kheish daemon HTTP API and event streams.
- [Debugging and recovery](https://docs.kheish.ai/operators/debugging-and-recovery.md): Evidence-first debugging, suspended runs, daemon restarts, and common operational pitfalls.
- [Deployment and hardening](https://docs.kheish.ai/operators/deployment-and-hardening.md): Practical deployment guidance for operating Kheish safely in production environments.
- [Docker and containers](https://docs.kheish.ai/operators/docker-and-containers.md): Running the Kheish daemon in Docker or container orchestration without weakening the runtime boundary.
- [MCP catalog](https://docs.kheish.ai/operators/mcp-catalog.md): Enable built-in MCP profiles and validate them on a real daemon.
- [OAuth](https://docs.kheish.ai/operators/oauth.md): Connect daemon-managed account auth and MCP OAuth accounts without exposing tokens to agents.
- [Operating playbooks and flows](https://docs.kheish.ai/operators/playbooks-and-flows.md): Creating, publishing, starting, inspecting, and cancelling Playbook-backed Flows on a real daemon.
- [Production runbooks](https://docs.kheish.ai/operators/production-runbooks.md): Production-facing runbooks for TLS proxying, backup and restore, key rotation, incidents, and operational smoke checks.
- [Running the daemon](https://docs.kheish.ai/operators/running-the-daemon.md): Building, starting, and operating a Kheish daemon safely.
- [Runtime configuration](https://docs.kheish.ai/operators/runtime-configuration.md): Live runtime reconfiguration for models, permissions, hooks, system prompt, and debug capture.
- [Security and auth](https://docs.kheish.ai/operators/security-and-auth.md): Control-plane authentication, brokered runtime auth, connector auth, signed external audit, and deployment boundaries.
- [Assets API](https://docs.kheish.ai/reference/assets-api.md): Import, inspect, and reuse daemon-owned assets for document and image inputs.
- [Boards API](https://docs.kheish.ai/reference/boards-api.md): Durable board records and immutable rendered board revisions.
- [Built-in tools](https://docs.kheish.ai/reference/built-in-tools.md): The default coding and web tools available in a standard Kheish runtime.
- [Channels API](https://docs.kheish.ai/reference/channels-api.md): Durable shared channels, public messages, reactions, public turn leases, stimuli, and canonical thread-work state.
- [Connector configuration](https://docs.kheish.ai/reference/connector-configuration.md): TOML structure and field-level guidance for HTTP, Slack, Telegram, and external connectors.
- [Connectors API](https://docs.kheish.ai/reference/connectors-api.md): Runtime connector CRUD, ingress payloads, auth behavior, session materialization, and reply-target defaults.
- [Control plane API](https://docs.kheish.ai/reference/control-plane-api.md): Base conventions, auth boundaries, and the reference map for the daemon HTTP API.
- [External connectors protocol](https://docs.kheish.ai/reference/external-connectors-protocol.md): Runtime-managed sidecar configuration, current ingress semantics, batch submission, and delivery compatibility for external connectors.
- [Hook reference](https://docs.kheish.ai/reference/hook-reference.md): Hook events, executor types, and the data model used by daemon-configured hooks.
- [Learnings API](https://docs.kheish.ai/reference/learnings-api.md): Candidate review, automatic capture and publication, durable learning records, and promoted procedural-skill endpoints.
- [Observations and derivations API](https://docs.kheish.ai/reference/observations-and-derivations-api.md): Observation sources, source-scoped uploads, materialization into runs, and deterministic asset derivations.
- [Personas API](https://docs.kheish.ai/reference/personas-api.md): Daemon-managed persona records, session persona bindings, and the binding rules that keep sessions stable.
- [Playbooks and flows API](https://docs.kheish.ai/reference/playbooks-and-flows-api.md): HTTP API for immutable Playbook versions and Flow projections over normal Kheish runs.
- [Projects API](https://docs.kheish.ai/reference/projects-api.md): Daemon-owned projects, project members, linked channels, project tasks, and project-task start operations.
- [Questions and approvals API](https://docs.kheish.ai/reference/questions-and-approvals-api.md): Resume suspended runs after permission approvals or structured user questions.
- [Runtime API](https://docs.kheish.ai/reference/runtime-api.md): Daemon-wide capabilities, routing, secrets, connectors, hooks, permission mode, debug level, and live events.
- [Sessions and runs API](https://docs.kheish.ai/reference/sessions-and-runs-api.md): Session lifecycle, session-scoped defaults, inline vs detached execution, and run inspection.
- [Skill format](https://docs.kheish.ai/reference/skill-format.md): On-disk layout, frontmatter fields, runtime metadata, and precedence rules for deployable Kheish skills.
- [Skills API](https://docs.kheish.ai/reference/skills-api.md): Skill catalog endpoints and the runtime surfaces used to inspect and activate reusable skills.
- [Tasks, schedules, and agents API](https://docs.kheish.ai/reference/tasks-schedules-agents-api.md): Session-task inspection, durable schedules, sidechain creation, and mailbox delivery.
- [Hooks](https://docs.kheish.ai/runtime/hooks.md): Lifecycle interception, policy enforcement, and runtime augmentation through daemon-configured hooks.
- [MCP](https://docs.kheish.ai/runtime/mcp.md): Model Context Protocol integration and how MCP-backed tools are surfaced into the runtime.
- [OpenRouter](https://docs.kheish.ai/runtime/openrouter.md): Configure OpenRouter as a first-class route, understand the supported daemon surfaces, and use route-aware selectors safely.
- [Permissions](https://docs.kheish.ai/runtime/permissions.md): Permission modes, rule scopes, approval behavior, and auditability for tool execution.
- [Providers and routing](https://docs.kheish.ai/runtime/providers-and-routing.md): Named routes, route files, selector grammar, capability-aware routing, and route pinning.
- [Skills](https://docs.kheish.ai/runtime/skills.md): Reusable instruction bundles, discovery roots, session-scoped activation, and explicit skill usage.
- [Tools](https://docs.kheish.ai/runtime/tools.md): Built-in coding and web tools, daemon control tools, and how tool execution fits into the runtime.