Now in Beta · OpenAI-Compatible · Multi-Tenant

Docker for AI Agents

An open spec, runtime, and toolchain for packaging AI agents as portable, observable, deployable artifacts.

Define agents in YAML. Bundle knowledge bases with vector embeddings. Deploy to any infrastructure. Observe every request.

Get started free View on GitHub

Step 1

Define your agent

Write a YAML spec with your system prompt, skills, knowledge base reference, and merge policies. It's your agent's portable definition.

agent.yaml

apiVersion: arachne.ai/v0
kind: Agent
metadata:
  name: support-agent
spec:
  model: gpt-4o
  systemPrompt: |
    You are a helpful support agent.
    Answer using the knowledge base.
  knowledgeBaseRef: product-docs

Step 2

Build and push

Weave your agent into a deployable artifact with embedded knowledge base. Push to any Arachne registry.

Terminal

$ arachne weave agent.yaml
  Weaving support-agent...
  Chunking 42 documents (650 tokens, 120 overlap)
  Embedding 186 chunks via text-embedding-3-small
  Packaged support-agent.orb (2.4 MB)

$ arachne push support-agent.orb
  Pushed myorg/support-agent:latest

Step 3

Deploy and observe

Deploy to any environment. Every request is traced with latency, token usage, and RAG retrieval metrics. Zero additional code required.

Terminal

$ arachne deploy myorg/support-agent:latest --env production
  Deployed to production

$ curl https://api.arachne-ai.com/v1/chat/completions \
    -H "Authorization: Bearer loom_sk_..." \
    -d '{"model":"gpt-4o","messages":[...]}'

Why Arachne?

Without Arachne

✗ Glue code for every LLM provider
✗ Separate vector DB setup and management
✗ Custom observability pipeline
✗ Conversation state managed in app code
✗ Agent configs scattered across services
✗ No standard way to package or deploy

With Arachne

✓ One API for OpenAI, Azure, Ollama
✓ Embedded vector search with pgvector
✓ Tracing built into the gateway
✓ Conversation memory as a runtime feature
✓ Portable YAML agent specs
✓ arachne weave → push → deploy

What the Runtime Gives You

Everything you need to ship AI agents

🧠

Conversation Memory

Automatic conversation history with token-aware summarization. Enable per-agent with one flag. History persists across requests with partition support.

📚

RAG Knowledge Bases

Upload documents, auto-chunk and embed them, then attach to any agent. Vector search retrieves relevant context for every request automatically.

🔍

Full Observability

Every request traced with latency, token usage, model, and provider. Fire-and-forget recording with zero gateway overhead. All data encrypted at rest.

🔀

Multi-Provider Routing

OpenAI, Azure OpenAI, and Ollama through a single API. Switch providers without changing client code. Per-agent or per-tenant configuration.

📦

Portable Agent Spec

Define agents in YAML with system prompts, skills, merge policies, and knowledge base refs. Weave into artifacts. Push to any registry.

👥

Multi-Tenant by Default

Built-in tenant isolation with subtenant hierarchies, per-tenant encryption, role-based access, and invite-based onboarding.