OpenClaw — AI Capability Manifest

Identity

OpenClaw is an infrastructure generation platform that produces production-ready Docker Compose stacks from a list of services, skill packs, and configuration options. It supports 95+ Docker services, 10+ presets, and 10+ skill packs covering AI, automation, monitoring, media, security, and more.

Connection Methods

MCP Server (Recommended for AI Agents)

{
  "mcpServers": {
    "better-openclaw": {
      "command": "npx",
      "args": ["-y", "@better-openclaw/mcp"]
    }
  }
}

Transport: stdio. Compatible with Claude Desktop, Cursor, Claude Code, and any MCP client.

REST API

Base URL: http://localhost:3456/api/v1

OpenAPI spec: GET /openapi.json
Swagger docs: GET /docs

CLI

npx create-better-openclaw generate --preset researcher

---

Available Tools

Discovery Tools

list-services

List all available Docker services. Optionally filter by category or maturity.

• Input: { category?: string, maturity?: "stable" | "beta" | "experimental" }
• Output: Array of { id, name, description, category, maturity, tags }

get-service

Get the complete definition of a service by ID.

• Input: { id: string }
• Output: Full ServiceDefinition (image, ports, env vars, volumes, dependencies, healthcheck, resource limits)

search-services

Keyword search across service names, descriptions, tags, and categories.

• Input: { query: string, limit?: number }
• Output: Scored results { id, name, description, category, score }

suggest-services

Suggest services from a natural language description of what to build.

• Input: { description: string, limit?: number }
• Output: Scored suggestions with match reasons

list-presets

List all curated presets (service bundles for common use cases).

• Input: {}
• Output: Array of { id, name, description, services[], estimatedMemoryMB }

get-preset

Get preset details with fully resolved dependency tree.

• Input: { id: string }
• Output: Preset definition + resolved service list with transitive dependencies

list-skill-packs

List skill packs, optionally filtered by service compatibility.

• Input: { serviceIds?: string[] }
• Output: Array of { id, name, description, requiredServices[], skills[], tags }

Composition Tools

resolve-dependencies

Resolve service dependencies, detect conflicts, and estimate memory.

• Input: { services: string[], skillPacks?: string[], proxy?, gpu?, platform?, monitoring? }
• Output: { isValid, services[], addedDependencies[], errors[], warnings[], estimatedMemoryMB }

validate-stack

Full validation pipeline: resolve -> compose -> validate (ports, volumes, YAML, DAG).

• Input: { services: string[], skillPacks?, proxy?, domain?, gpu?, platform?, monitoring? }
• Output: { valid, errors[], warnings[], serviceCount, estimatedMemoryMB }

generate-stack

Generate a complete Docker Compose stack. Returns all files as a JSON map.

• Input: { projectName: string, services: string[], skillPacks?, proxy?, domain?, gpu?, platform?, monitoring?, generateSecrets? }
• Output: { files: { [filename]: content }, metadata, fileList }

---

Data Model

Service IDs

Kebab-case identifiers. Examples: postgresql, redis, n8n, ollama, caddy, grafana, qdrant, meilisearch

Service Categories

ai, ai-platform, analytics, automation, browser, coding-agent, communication, database, desktop, dev-tools, knowledge, media, monitoring, proxy, search, security, social-media, storage, streaming, vector-db

Presets

minimal, creator, researcher, devops, full, content-creator, ai-playground, ai-powerhouse, coding-team, lasuite-meet

Skill Packs

video-creator, research-agent, social-media, dev-ops, knowledge-base, local-ai, content-creator, ai-playground, coding-team, knowledge-hub

GenerationInput Shape

{
  "projectName": "my-stack",
  "services": ["postgresql", "redis", "n8n"],
  "skillPacks": ["dev-ops"],
  "proxy": "none | caddy | traefik",
  "domain": "example.com",
  "gpu": false,
  "platform": "linux/amd64 | linux/arm64",
  "monitoring": false,
  "generateSecrets": true
}

---

MCP Resources

URI	Description
openclaw://services	Full service catalog with summaries
openclaw://presets	All preset definitions
openclaw://skills	All skill pack definitions

---

Example Agent Workflows

1. User wants a research stack

1. list-presets → find "researcher" preset
2. get-preset({ id: "researcher" }) → see resolved services and memory estimate
3. generate-stack({ projectName: "my-research", services: [...preset.services], proxy: "caddy", domain: "research.local" })
4. Return docker-compose.yml and .env to user

2. User asks "what can I use for AI?"

1. list-services({ category: "ai" }) → get all AI services
2. list-services({ category: "ai-platform" }) → get AI platform services
3. list-skill-packs() → find AI-related skill packs
4. suggest-services({ description: "AI chatbot with document search" }) → get recommendations
5. Present options to user with descriptions and memory estimates

3. User wants to add monitoring to their services

1. resolve-dependencies({ services: ["redis", "postgresql", "n8n"], monitoring: true })
   → see which monitoring services get auto-added (grafana, prometheus)
2. validate-stack({ services: [...resolved], monitoring: true })
   → verify no conflicts
3. generate-stack({ projectName: "my-app", services: [...], monitoring: true })

4. Build a custom stack from scratch

1. suggest-services({ description: "video processing with notifications and storage" })
   → get suggested services
2. resolve-dependencies({ services: ["ffmpeg", "minio", "ntfy"] })
   → check dependencies, get memory estimate
3. validate-stack({ services: [...resolved] })
   → verify everything works together
4. generate-stack({ projectName: "video-pipeline", services: [...], generateSecrets: true })
   → get all deployment files

---

Output Files

A generated stack typically includes:

• docker-compose.yml — Main compose file
• docker-compose.*.yml — Profile-specific compose files (ai, monitoring, tools, etc.)
• .env / .env.example — Environment variables with generated secrets
• README.md — Setup instructions
• scripts/ — Helper scripts (start, stop, backup, etc.)
• Caddyfile or traefik/ — Reverse proxy configuration (if proxy selected)
• grafana/ / prometheus/ — Monitoring configs (if monitoring enabled)
• init-db.sql — PostgreSQL initialization (if PostgreSQL used)
• skills/ — Skill documentation with connection details