# Documentation

- [Agents as Functions](/docs/agents/agents-as-functions): AEGIS agents are stateless, typed, invokable units — like serverless functions backed by language models. This guide explains the mental model and shows rich examples across classification, extraction, transformation, and validation use cases.
- [Configuring Tools](/docs/agents/configuring-tools): How to enable cmd.run subprocess execution and register external MCP tool servers in your node configuration and agent manifests.
- [Creating Agents via the UI](/docs/agents/creating-agents-ui): Step-by-step guide to creating, editing, and deleting agents using the Zaru web interface.
- [Custom Runtime Agents (Advanced)](/docs/agents/custom-runtime-agents): Advanced guide for building a custom container image and bootstrap script when manifest-only agents are not enough.
- [Deploying & Running Agents](/docs/agents/deploying-agents): Deploy, inspect, execute, and monitor agents with the current AEGIS CLI.
- [Aegis Dispatch Protocol](/docs/agents/dispatch-protocol): Runtime-agnostic in-container command execution via bootstrap.py.
- [The Execution Loop](/docs/agents/execution): How the 100monkeys outer loop, inner loop, and Dispatch Protocol work together to run an agent execution.
- [Importing Community Skills](/docs/agents/importing-skills): Use the built-in skill-import workflow to convert community SKILL.md definitions into AEGIS agent baselines
- [Local & Integration Testing](/docs/agents/local-testing): Build, run, and verify AEGIS locally — from a quick CLI smoke test to a full Docker-based Temporal workflow integration.
- [What are Agents?](/docs/agents/overview): The agent manifest format, lifecycle states, runtime selection, and BYOLLM model alias system.
- [Tool-Call Judging](/docs/agents/tool-call-judging): How AEGIS performs inner-loop semantic pre-execution review for tool calls, and how operators configure or bypass it safely.
- [Configuring Agent Validation](/docs/agents/validation-guide): Validator types, gradient scoring, threshold configuration, inner-loop tool-call judging, and chaining multiple validators.
- [Validation](/docs/agents/validation): How AEGIS uses gradient scoring, judge agents, and consensus to determine whether an iteration's output is acceptable.
- [Writing Your First Agent](/docs/agents/writing-agents): Step-by-step guide to creating your first agent with a declarative manifest.
- [Event Bus](/docs/architecture/event-bus): The AEGIS domain event bus — tokio::sync::broadcast, DomainEvent catalog, subscriber patterns, and Phase 1 constraints.
- [Execution Engine](/docs/architecture/execution-engine): Deep dive into the ExecutionSupervisor outer loop, InnerLoopService, and the Aegis Dispatch Protocol.
- [Identity and Secrets](/docs/architecture/identity-and-secrets): The Keymaster Pattern, OAuth federation, and centralized secrets management.
- [Architecture Overview](/docs/architecture/overview): Component map, bounded contexts, data flow, and technology stack for the AEGIS orchestrator.
- [Stimulus-Response Routing](/docs/architecture/stimulus-response): Architecture deep-dive into the two-stage hybrid routing pipeline for external event ingestion (BC-8).
- [Storage Backends](/docs/architecture/storage-backends): Comprehensive guide to the storage backends supported by AEGIS, including SeaweedFS, OpenDAL, Local Host mounts, and SEAL remote volumes.
- [Storage Gateway](/docs/architecture/storage-gateway): AegisFSAL architecture, FUSE and NFS transports, FileHandle structure, UID/GID squashing, path canonicalization, and SeaweedFS integration.
- [MCP Tool Routing](/docs/architecture/tool-routing): The three tool routing paths — FSAL host file operations, SEAL External host MCP servers, and Dispatch Protocol in-container execution.
- [SEAL Tooling Gateway](/docs/architecture/tooling-gateway): Where the SEAL Tooling Gateway sits in the AEGIS stack and a pointer to its dedicated documentation section.
- [Workflow Engine](/docs/architecture/workflow-engine): Architecture of the AEGIS workflow FSM, Temporal integration, Blackboard system, and workflow execution lifecycle.
- [Agents](/docs/concepts/agents): Declarative units of work that run in isolated runtimes and refine their own output.
- [Cluster and Nodes](/docs/concepts/cluster-and-nodes): How an AEGIS cluster is composed of nodes with roles, and how workloads are routed between them.
- [Cortex (Learning & Memory) Integration](/docs/concepts/cortex-integration): How the AEGIS orchestrator connects to the central Cortex memory graph and operates in memoryless mode.
- [Edge Mode Overview](/docs/concepts/edge-mode-overview): How AEGIS Edge Mode lets you install a daemon on your own host, bind it to your tenant, and route MCP tool calls through it to perform local operations.
- [Edge Security](/docs/concepts/edge-security): The two-envelope SEAL pattern, enrollment token lifecycle, local SecurityContext enforcement, key rotation, and tenant-isolation guarantees for AEGIS edge daemons.
- [Executions](/docs/concepts/executions): A single run of an agent, including its iterative refinement loop.
- [Fleet Operations](/docs/concepts/fleet-operations): Targeting, dispatch policies, failure policies, and per-node streaming results for multi-host edge fleet operations.
- [Observability](/docs/concepts/observability): Platform-wide structured logging, metrics, and tracing built on OpenTelemetry.
- [SEAL Gateway](/docs/concepts/seal-gateway): A signed, attested boundary between agents and external APIs that compresses multi-step REST calls into single tool invocations.
- [Secrets and Identity](/docs/concepts/secrets-and-identity): How OpenBao stores credentials centrally and Keycloak handles authentication, so agents never see raw secrets.
- [Security Model](/docs/concepts/security-model): The two-layer AEGIS security model — infrastructure policy enforcement and SEAL protocol-level security.
- [Stimulus-Response Routing](/docs/concepts/stimulus-response): How external events trigger workflow executions through the two-stage routing pipeline.
- [Storage Gateway](/docs/concepts/storage-gateway): How agents see a normal POSIX filesystem while the orchestrator transparently enforces security and routes I/O to the right backend.
- [Swarms](/docs/concepts/swarms): How a parent agent spawns child agents to coordinate multi-agent work under a shared execution context.
- [Tenants](/docs/concepts/tenants): The isolation boundary for users, secrets, executions, and routing in AEGIS.
- [Tools](/docs/concepts/tools): MCP-style capabilities that agents call to read the world and act on it.
- [Workflows](/docs/concepts/workflows): Declarative state machines that coordinate agents, containers, and human approvals.
- [Zaru](/docs/concepts/zaru): The consumer-facing product that puts AEGIS in front of real users at myzaru.com.
- [Backup & Restore](/docs/deployment/backup-restore): Backup and restore procedures for AEGIS stateful services — PostgreSQL, OpenBao, SeaweedFS, and coordinated backup strategies.
- [Caddy Edge Proxy](/docs/deployment/caddy-edge-proxy): Reverse proxy configuration with Caddy for production AEGIS deployments — automatic TLS, domain routing, and Cloudflare DNS-01.
- [Configuration Reference](/docs/deployment/configuration): Full annotated aegis-config.yaml with all supported keys, types, defaults, and descriptions.
- [Configuring Security Contexts](/docs/deployment/configuring-security-contexts): Define named SecurityContext boundaries to control which MCP tools agents can invoke, with path, command, domain, and rate limit constraints.
- [Configuring Storage Volumes](/docs/deployment/configuring-storage): Declaring ephemeral and persistent volumes, volume mounts, access modes, quotas, and managing volumes via CLI.
- [Configuring Webhooks](/docs/deployment/configuring-webhooks): Connect external systems to AEGIS workflows using HMAC-authenticated webhooks and direct-route registration.
- [Container Registry & Image Management](/docs/deployment/container-registry): How AEGIS discovers, pulls, caches, and authenticates container images for standard and custom runtimes — including ImagePullPolicy, private registry credentials, failure scenarios, and pre-caching for airgapped environments.
- [Disaster Recovery](/docs/deployment/disaster-recovery): Recovery runbooks for AEGIS platform failure scenarios — pod failures, database corruption, secrets engine recovery, and full environment rebuild.
- [Docker Deployment](/docs/deployment/docker): Docker Engine setup, agent container lifecycle, bollard integration, NFS volume mounting, and systemd service configuration.
- [Edge Daemon Installation](/docs/deployment/edge-daemon-installation): Install the AEGIS edge daemon on Linux (systemd user unit), macOS (launchd plist), or Windows (NSSM service), with the hardened service templates.
- [Edge Operational Patterns](/docs/deployment/edge-operational-patterns): Day-2 operations for AEGIS edge daemons — monitoring health, log and trace collection, daemon updates, decommissioning, and incident response.
- [Edge Relay Deployment](/docs/deployment/edge-relay-deployment): Deploy the Relay Coordinator service for self-hosted AEGIS — pod manifest, ingress, h2c reverse-proxy configuration, certificates, and bootstrap.
- [Firecracker Deployment](/docs/deployment/firecracker): KVM requirements, AgentRuntime trait abstraction, isolation model, jailer hardening, and virtio-fs storage transport.
- [IAM & Identity Federation](/docs/deployment/iam): Keycloak realms, AegisRole claims, M2M service accounts, SAML federation, JWKS cache, and GrpcIamAuthInterceptor.
- [Infrastructure Requirements](/docs/deployment/infrastructure): Dependency matrix, hardware requirements, network topology, pod architecture, and the local development stack.
- [Configuring LLM Providers](/docs/deployment/llm-providers): The BYOLLM model alias system — configuring OpenAI, Anthropic, and Ollama providers and mapping them to agent-facing aliases.
- [Multi-Node Deployment](/docs/deployment/multi-node): Distribute AEGIS across multiple machines for production deployments using the gRPC cluster protocol.
- [Multi-Tenancy Configuration](/docs/deployment/multi-tenancy): Configure and isolate multiple tenants using Keycloak realms and OpenBao namespaces.
- [Observability](/docs/deployment/observability): Observability stack (Jaeger, Prometheus, Grafana, Loki), structured logging, OTLP export, alert rules, and metrics for AEGIS deployments.
- [Pod Architecture](/docs/deployment/pod-architecture): Complete pod topology for AEGIS platform deployment — containers, ports, health checks, persistent volumes, and inter-pod communication.
- [Podman Deployment](/docs/deployment/podman): Production platform deployment with Podman pods, deployment profiles, Makefile automation, and Podman as the rootless agent container runtime.
- [Production Hardening](/docs/deployment/production-hardening): Security checklist and hardening guide for production AEGIS deployments — TLS, secrets rotation, resource limits, network segmentation, and access control.
- [Deployment Profiles](/docs/deployment/profiles): Profile-based deployment for AEGIS — minimal, development, and full profiles with pod composition, use cases, and environment configuration.
- [SeaweedFS Storage](/docs/deployment/seaweedfs): Deploying SeaweedFS as the AEGIS volume storage backend — single-node dev setup, production HA topology, replication, S3 gateway, and health checks.
- [Secrets Management](/docs/deployment/secrets): OpenBao Keymaster Pattern, AppRole authentication, KV and dynamic secrets, Transit Engine, and multi-tenant namespaces.
- [Temporal Deployment](/docs/deployment/temporal): Running Temporal Server, the TypeScript workflow worker, and PostgreSQL for durable workflow execution in AEGIS.
- [Upgrade Procedures](/docs/deployment/upgrade-procedures): How to upgrade AEGIS platform components — image updates, database migrations, rolling upgrades, version compatibility, and rollback procedures.
- [API Explorer](/docs/gateway/api-explorer): A thin proxy that issues a single HTTP call against a registered API spec and returns only the JSONPath-selected fields — designed to keep agent contexts small.
- [Authentication](/docs/gateway/authentication): How operators and agents authenticate to the SEAL Gateway — operator JWTs for the control plane and SEAL envelopes for tool invocation.
- [Authoring Workflows](/docs/gateway/authoring-workflows): Chain operations from a registered API spec into a single named tool — Handlebars body templates, JSONPath extractors, error policies, and a worked Terraform Cloud example.
- [Concepts](/docs/gateway/concepts): The eleven concepts the rest of the SEAL Gateway documentation assumes — envelopes, tokens, contexts, credential paths, and the lifecycle of a tool call.
- [Configuration](/docs/gateway/configuration): Complete configuration reference — file discovery, YAML schema, environment variable overrides, precedence, and worked examples.
- [Credential Resolution](/docs/gateway/credential-resolution): How the SEAL Gateway resolves upstream credentials at invocation time — the five resolution strategies, their configuration, and their failure modes.
- [Deployment](/docs/gateway/deployment): Production runbook for the SEAL Gateway — container image, Docker Compose, Kubernetes, systemd, storage choice, networking, upgrades, and backups.
- [Ephemeral CLI Tools](/docs/gateway/ephemeral-cli-tools): Register Docker images the gateway runs once per invocation — subcommand allowlists, the optional semantic judge, and worked examples for terraform, kubectl, gh, and aws.
- [gRPC API](/docs/gateway/grpc-api): gRPC complement to the REST control plane — ToolWorkflowService, GatewayInvocationService, client generation in Go, Python, and TypeScript.
- [Integration with AEGIS](/docs/gateway/integration-with-aegis): How the SEAL Gateway is wired into the AEGIS Orchestrator, including session provisioning, security context coordination, the empty native tool catalog, and co-deployment topology.
- [Management API (REST)](/docs/gateway/management-api): Complete REST reference for the SEAL Gateway control plane — API specs, workflows, CLI tools, SEAL sessions, security contexts, and unified tool catalog.
- [Observability](/docs/gateway/observability): Structured logs, audit events, and the telemetry surfaces the SEAL Gateway exposes today — plus the gaps you need to know about.
- [Overview](/docs/gateway/overview): A standalone, signed-envelope tooling gateway that puts policy, credentials, and audit between automated clients and arbitrary upstream tools.
- [Quickstart](/docs/gateway/quickstart): From zero to a SEAL-signed tool call in ten minutes — standalone, no AEGIS Orchestrator required.
- [Registering API Specs](/docs/gateway/registering-api-specs): Register OpenAPI 3.0 documents with the SEAL Gateway so workflows and the API Explorer can call HTTP operations on your behalf.
- [SEAL Protocol](/docs/gateway/seal-protocol): Signed Envelope Attestation Layer — wire format, signature construction, per-call authorization, replay prevention, error codes, and client implementation guidance.
- [Security Contexts](/docs/gateway/security-contexts): Define named, server-side capability policies that gate every SEAL tool call by tool name, path, command, subcommand, domain, and output size.
- [Troubleshooting](/docs/gateway/troubleshooting): SEAL Gateway runbook — symptom-driven diagnosis and remediation for the failure modes operators actually hit in production.
- [Getting Started](/docs/getting-started): Install AEGIS, start the local stack, and run your first agent execution.
- [Authentication](/docs/guides/authentication): Authenticate the AEGIS CLI with your environment using aegis auth, and generate API keys for programmatic access.
- [Edge Fleet Operations](/docs/guides/edge-fleet-operations): Run tools across many edge hosts — rolling restarts, ad-hoc commands, redeploys — read streamed per-node results, and cancel runs in flight.
- [Edge Key Rotation](/docs/guides/edge-key-rotation): When and how to rotate edge daemon keys, the dual-signature protocol, the overlap window, and fleet-wide rotation for blast-radius events.
- [Management via MCP](/docs/guides/management-via-mcp): How to use built-in management tools to automate agent and workflow lifecycles.
- [Troubleshooting](/docs/guides/troubleshooting): Diagnostic commands, common failure patterns, service dependency chains, and log analysis for AEGIS platform deployments.
- [Welcome to AEGIS](/docs): AEGIS is an open-source orchestrator for running autonomous AI agents in isolated, policy-enforced runtimes.
- [Agent Manifest Reference](/docs/reference/agent-manifest): Complete specification for the AgentManifest YAML format (v1.0) — schema, field definitions, examples, and validation configuration.
- [Built-in Tools](/docs/reference/builtin-tools): Reference documentation for the built-in MCP tools provided by the AEGIS Orchestrator
- [CLI Capability Matrix](/docs/reference/cli-capability-matrix): Current CLI command coverage, MCP tool equivalents, and API-only surfaces in the AEGIS orchestrator.
- [CLI Reference](/docs/reference/cli): Reference for the currently implemented aegis CLI commands and flags.
- [Edge CLI Reference](/docs/reference/edge-cli): Full reference for the aegis edge CLI family — enrollment, status, listing, tagging, groups, fleet operations, key and token management, and exit codes.
- [Edge Config Reference](/docs/reference/edge-config): Reference for the cluster.edge configuration block, file layout, permission expectations, auto-generated template fields, and the bootstrap prompt matrix.
- [Edge REST API](/docs/reference/edge-rest-api): HTTP reference for the /v1/edge/* surface — enrollment tokens, hosts, tags, groups, fleet preview, fleet invoke, fleet cancel, and run history.
- [gRPC API Reference](/docs/reference/grpc-api): aegis.runtime.v1 AegisRuntime service — the Temporal-worker-facing gRPC transport.
- [Management Tools](/docs/reference/management-tools): Built-in tools for managing AEGIS agents, workflows, and schemas via MCP.
- [MCP Tool Reference](/docs/reference/mcp-tool-reference): Complete input and output schemas for all AEGIS MCP tools — built-in orchestrator tools and the Zaru-only zaru.init, zaru.mode, zaru.docs, zaru.memory.get, and zaru.memory.set tools — with availability by context and Zaru conversation mode.
- [Metrics Reference](/docs/reference/metrics): Complete catalog of Prometheus metrics exposed by the AEGIS orchestrator, including names, types, labels, and descriptions.
- [Node Configuration Reference](/docs/reference/node-config): Complete specification for the NodeConfig YAML format (v1.0) — schema, field definitions, credential resolution, model alias system, and example configurations.
- [Relay gRPC API](/docs/reference/relay-grpc-api): gRPC reference for the daemon-facing extensions to NodeClusterService — ConnectEdge bidirectional stream, RotateEdgeKey, EdgeEvent and EdgeCommand messages.
- [Releases](/docs/reference/releases): The 1.0.1-beta release of every published AEGIS and Zaru package and image, and where each is published.
- [REST API Reference](/docs/reference/rest-api): Full HTTP API surface of the AEGIS Orchestrator — all routes, request/response schemas, and authentication.
- [Python SDK](/docs/reference/sdk-python): AegisClient async HTTP client and bootstrap.py Dispatch Protocol types for Python agent authors.
- [TypeScript SDK](/docs/reference/sdk-typescript): AegisClient for TypeScript and Node.js — agents, executions, workflows, volumes, credentials, SEAL, and more.
- [Standard Runtime Registry](/docs/reference/standard-runtime-registry): The complete language-version-to-image mapping table for AEGIS StandardRuntime — supported versions, base images, pre-installed toolchains, and the version support lifecycle.
- [Template Syntax](/docs/reference/template-syntax): Handlebars template variables and helpers available in agent prompt templates and workflow Blackboard templates.
- [Workflow Manifest Reference](/docs/reference/workflow-manifest): Complete specification for the WorkflowManifest YAML format — schema, field definitions, all seven state kinds including ContainerRun for CI/CD steps, Subworkflow for workflow composition, consensus strategies, and complete examples.
- [Building Multi-Agent Swarms](/docs/workflows/building-swarms): How to spawn child agents, pass messages, use resource locks, and manage swarm lifecycle.
- [Building Workflows](/docs/workflows/building-workflows): Step-by-step guide to writing, deploying, and running multi-agent workflow FSMs with the Forge reference pattern.
- [Building CI/CD Pipelines](/docs/workflows/cicd-pipelines): Mix agent steps and container-based CI/CD steps — build, test, lint, and deploy — in a single unified AEGIS workflow.
- [Creating Workflows via the UI](/docs/workflows/creating-workflows-ui): Step-by-step guide to creating, editing, and managing workflows using the Zaru web interface, including scope management and FSM visualization.
- [Human Approvals](/docs/workflows/human-approvals): Pause workflow execution for human review, approval, or rejection using the built-in approval gate.
- [What are Workflows?](/docs/workflows/overview): Declarative finite state machine workflows backed by Temporal. State types including ContainerRun for CI/CD steps, Blackboard context, transition rules, and the Forge reference pattern.
- [Schedules](/docs/workflows/schedules): Start an agent or a workflow later, once or on a recurrence, as the person who made the schedule.
- [Swarms](/docs/workflows/swarms): Multi-agent swarm coordination — spawning child executions, inter-agent messaging, resource locking, and cascade cancellation.
- [Workflow Visualization](/docs/workflows/workflow-visualization): How to read the FSM state machine graph, including color coding, initial state highlighting, and transition labels.
- [Zaru CLI Command Reference](/docs/zaru-cli/commands): Every command, flag and in-session command of the zaru program, and its exit codes.
- [Zaru CLI Configuration](/docs/zaru-cli/configuration): Where the zaru program reads its settings, every configuration key it reads, and the project file zaru.toml.
- [Installing the Zaru CLI](/docs/zaru-cli/installation): Build the zaru program from source, choose a model, and store a provider key.
- [Zaru CLI Overview](/docs/zaru-cli/overview): The zaru command-line program, an AI assistant that works in your terminal. Pre-alpha, built from source.
- [Zaru CLI Permissions](/docs/zaru-cli/permissions): How the zaru program asks before it acts, what each answer does, the permission modes, the allowlist, and approving a project's validators and commands.
- [Running a Task with the Zaru CLI](/docs/zaru-cli/running-a-task): What the zaru program does and shows while a task runs, the tools the model can call, the context window, validators, and sessions.
- [Zaru CLI, Zaru Web and AEGIS](/docs/zaru-cli/zaru-web-and-aegis): What the zaru command-line program shares with Zaru Web and with the AEGIS orchestrator at this commit, and what it does not.
- [API Keys](/docs/zaru/api-keys): Create and manage API keys for programmatic access to the AEGIS platform via the MCP server and SDKs.
- [Calendar](/docs/zaru/calendar): Connect a calendar account, by Google sign-in or a CalDAV server's address, and let Zaru read your events and, with your approval, change them.
- [Chat](/docs/zaru/chat): The main conversational interface at ask.myzaru.com — conversation modes, goals in Agentic mode, model selection, and the Glass Laboratory context panel.
- [Code Mode](/docs/zaru/code-mode): Build browser apps with live preview — vibe-code inside Zaru with no local dev environment.
- [Context Panel & Glass Laboratory](/docs/zaru/context-panel): The real-time context panel for browsing agents, workflows, tools, scripts, and monitoring execution progress.
- [Zaru Client Admin Console](/docs/zaru/control-plane): The unified operator and consumer interface for managing the AEGIS orchestrator, delivered through the Zaru client.
- [Edge Host Setup](/docs/zaru/edge-host-setup): Add an edge host to your Zaru tenant via the Vault → Edge Hosts UI, install and enroll the AEGIS edge daemon on Linux, macOS, or Windows, and confirm connectivity.
- [Edge Tag and Group Management](/docs/zaru/edge-tag-and-group-management): How to assign tags to edge hosts, build reusable groups with saved selectors, and preview which hosts a target hits before you fan out.
- [Running Code with Execute Mode](/docs/zaru/execute-mode): Use Execute mode in Zaru to describe a computation in plain English, provide typed inputs, and get back a result — no workflow authoring required.
- [Beta Invites](/docs/zaru/invites): How the invite system works during the Zaru beta — sending, redeeming, and managing invite codes.
- [Live Mode](/docs/zaru/live-mode): Write and execute TypeScript with access to the AEGIS SDK directly in chat.
- [Connect Zaru to Any MCP Client](/docs/zaru/mcp-client-setup): Set up the Zaru MCP server in Claude Code, Claude Desktop, Cursor, Windsurf, VS Code Copilot, Gemini CLI, and other MCP-compatible tools.
- [MCP Tools](/docs/zaru/mcp-tools): The tools Zaru exposes to MCP clients, organized by conversation mode — what each tool does, when it is available, and how to call it.
- [Memory](/docs/zaru/memory): A single, evolving markdown document Zaru keeps about you so the next conversation builds on the last.
- [Overview](/docs/zaru/overview): How the Zaru consumer product interfaces with the AEGIS orchestrator.
- [Pricing & Plans](/docs/zaru/pricing): Zaru tier comparison — Free, Pro, Business, and Enterprise — with details on features, limits, and billing.
- [Profiles](/docs/zaru/profiles): A named set of your connections and the tools they may use, chosen for a conversation, a run or a schedule in one step.
- [The Architect](/docs/zaru/the-architect): Create, manage, and monitor agents and workflows from the Architect dashboard.
- [The Colony](/docs/zaru/the-colony): Manage your team, configure SSO, and handle subscription billing from The Colony.
- [The Workshop](/docs/zaru/the-workshop): Create and manage persistent storage volumes for your agent workloads.
- [The Vault](/docs/zaru/vault): Store and manage credentials that your agents use at runtime. Agents never see credential values directly — the platform injects them securely at execution time.