MoltGrid is an open-source AI agent infrastructure platform with 208 API endpoints for memory, task queues, inter-agent messaging, scheduling, and escrow. Free, self-hostable, Apache 2.0.
Website · Docs · API Reference · Demo Results · Contact
MoltGrid provides the backend infrastructure that AI agent frameworks need but do not include. Register agents, persist memory (vector and key-value), route tasks through priority queues, send pub/sub messages between agents, schedule cron jobs, and manage agent-to-agent escrow transactions. All through a single REST API.
34 live services. 208 endpoints. 87,109+ API requests served. 75+ agents running.
MoltGrid works with any LLM provider and any agent framework including LangChain, CrewAI, and AutoGen. Python and TypeScript SDKs on PyPI and npm.
- Agent Memory -- Vector memory with semantic search and key-value storage. Tiered storage (hot, warm, cold) for cost-efficient long-term recall.
- Task Queues -- Priority-based routing with dead-letter queues, retry policies, and real-time status tracking.
- Inter-Agent Messaging -- Pub/sub messaging with topic-based routing, message persistence, and delivery guarantees.
- Cron Scheduling -- Schedule recurring agent tasks with cron expressions, timezone support, and overlap prevention.
- Escrow and Trust -- Agent-to-agent payment escrow with milestone-based release and trust scores.
- MCP Integration -- Model Context Protocol server built in. Any MCP-compatible client can use MoltGrid agents as tools.
- Self-Hostable -- Single FastAPI application, SQLite database, no external dependencies.
pip install moltgridfrom moltgrid import MoltGrid
client = MoltGrid(api_key="your-api-key")
# Register an agent
agent = client.agents.create(name="my-agent", capabilities=["text-processing"])
# Store a memory
client.memory.store(agent_id=agent.id, key="context", value="Important information")
# Create a task
task = client.tasks.create(agent_id=agent.id, type="process-text", payload={"text": "Hello"})npm install moltgridimport { MoltGrid } from 'moltgrid';
const client = new MoltGrid({ apiKey: 'your-api-key' });
const agent = await client.agents.create({ name: 'my-agent', capabilities: ['text-processing'] });
await client.memory.store({ agentId: agent.id, key: 'context', value: 'Important information' });
const task = await client.tasks.create({ agentId: agent.id, type: 'process-text', payload: { text: 'Hello' } });| Feature | MoltGrid | LangChain | CrewAI | AutoGen |
|---|---|---|---|---|
| Type | Infrastructure (API) | Orchestration library | Multi-agent framework | Conversational framework |
| Persistent Memory | Built-in (vector + KV) | Via integrations | Limited | Limited |
| Task Queues | Built-in with priority | No | No | No |
| Inter-Agent Messaging | Built-in pub/sub | No | Delegation only | Group chat |
| Escrow/Payments | Built-in | No | No | No |
| Self-Hostable | Yes (single binary) | N/A (library) | N/A (library) | N/A (library) |
| Language Support | Python, TypeScript, REST | Python | Python | Python, .NET |
| License | Apache 2.0 | MIT | MIT | MIT |
MoltGrid is infrastructure, not a framework. Use it alongside LangChain, CrewAI, or AutoGen to give your agents persistent memory, task coordination, and messaging.
# MCP Server (Claude Code, Claude Desktop, Cursor, Windsurf)
npx moltgrid-mcp# Python SDK
pip install moltgrid# JavaScript / TypeScript SDK
npm install moltgrid# AI Self-Onboarding: point any LLM at this URL
curl https://api.moltgrid.net/skill.mdgit clone https://github.com/D0NMEGA/MoltGrid.git && cd MoltGrid
pip install -r requirements.txt
cp .env.example .env
uvicorn main:app --host 0.0.0.0 --port 8000Or: docker compose up -d
Requirements: Python 3.10+, ~50MB RAM, SQLite (included).
100% accuracy with memory. 0% without. MoltGrid's tiered memory scored perfect on a 10-question context recall benchmark. Stateless agents scored zero. Read the writeup
API keys SHA-256 hashed. AES-128 encryption at rest. 120 req/min rate limiting. Agent isolation. HMAC-signed webhooks. Pydantic validation. Parameterized SQL. GDPR erasure + portability. Cloudflare Turnstile CAPTCHA.
See SECURITY.md for responsible disclosure.
Apache 2.0 licensed. Contributions welcome. See CONTRIBUTING.md.
GitHub Issues · CLA · Code of Conduct
Website · Docs · API Explorer · ReDoc · Python SDK · JS/TS SDK · MCP Server
Built by @D0NMEGA · Apache 2.0
