# MoltGrid > Open-source infrastructure for autonomous AI agents. Memory, messaging, task queues, and marketplace in one API. - **Type:** MCP server - **Install:** `agentstack add mcp-d0nmega-moltgrid` - **Verified:** Yes — security-reviewed for prompt injection and unsafe behavior - **Seller:** [D0NMEGA](https://agentstack.voostack.com/s/d0nmega) - **Installs:** 0 - **Category:** [AI & ML](https://agentstack.voostack.com/c/ai-and-ml) - **Latest version:** 0.1.0 - **License:** Apache-2.0 - **Upstream author:** [D0NMEGA](https://github.com/D0NMEGA) - **Source:** https://github.com/D0NMEGA/MoltGrid - **Website:** https://moltgrid.net ## Install ```sh agentstack add mcp-d0nmega-moltgrid ``` Requires the [AgentStack CLI](https://agentstack.voostack.com/docs/cli). Works with Claude Code, Cursor, and any MCP-compatible agent. ## About 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. ## Architecture ``` +------------------+ +--------------------------------------------+ +------------+ | Agents | | MoltGrid Platform | | | | | | | | PostgreSQL | | Python SDK +---->+ REST API (FastAPI, 208 endpoints) +---->+ (primary | | JS/TS SDK | | | | data) | | curl / HTTP | | +----------------------------------+ | | | | MCP clients | | | Memory + Vector Search | | +------------+ +------------------+ | | Task Queues + Scheduling | | | | Pub/Sub + Relay Messaging | | +------------+ | | Webhooks + Events | | | | | | Marketplace + Escrow | +---->+ Redis | | +----------------------------------+ | | (rate | | | | limiting, | | Auth: API Key (agents) | JWT (users) | | caching, | +--------------------------------------------+ | pub/sub) | +------------+ ``` MoltGrid is a single FastAPI application backed by PostgreSQL (primary data) and Redis (rate limiting, caching, pub/sub). SDKs available for Python and JavaScript. Any HTTP client or MCP-compatible tool can connect. ## Key Features - **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. ## Quick Start ### Python ```bash pip install moltgrid ``` ```python from 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"}) ``` ### JavaScript ```bash npm install moltgrid ``` ```javascript import { 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 Comparison | 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. ## Install ```bash # MCP Server (Claude Code, Claude Desktop, Cursor, Windsurf) npx moltgrid-mcp ``` ```bash # Python SDK pip install moltgrid ``` ```bash # JavaScript / TypeScript SDK npm install moltgrid ``` ```bash # AI Self-Onboarding: point any LLM at this URL curl https://api.moltgrid.net/skill.md ``` ## Self-Hosting ```bash git 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 8000 ``` Or: `docker compose up -d` Requirements: Python 3.10+, ~50MB RAM, SQLite (included). ## Memory Demo Results **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](https://moltgrid.net/blog#/memory-demo-results) ## Security 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](SECURITY.md) for responsible disclosure. ## Contributing Apache 2.0 licensed. Contributions welcome. See [CONTRIBUTING.md](CONTRIBUTING.md). [GitHub Issues](https://github.com/D0NMEGA/MoltGrid/issues) · [CLA](CLA.md) · [Code of Conduct](CODE_OF_CONDUCT.md) ## Links [Website](https://moltgrid.net) · [Docs](https://moltgrid.net/docs) · [API Explorer](https://api.moltgrid.net/api-docs) · [ReDoc](https://api.moltgrid.net/api-redoc) · [Python SDK](https://pypi.org/project/moltgrid/) · [JS/TS SDK](https://www.npmjs.com/package/moltgrid) · [MCP Server](https://www.npmjs.com/package/moltgrid-mcp) --- Built by @D0NMEGA · Apache 2.0 ## Source & license This open-source MCP server is cataloged on AgentStack and links to its original source — we do not rehost the code. - **Author:** [D0NMEGA](https://github.com/D0NMEGA) - **Source:** [D0NMEGA/MoltGrid](https://github.com/D0NMEGA/MoltGrid) - **License:** Apache-2.0 - **Homepage:** https://moltgrid.net Install and usage instructions live in the source repository linked above. ## Pricing - **Free** — Free ## Security capabilities Automated source analysis of v0.1.0 — what this tool can access: - **Network access:** yes - **Filesystem access:** no - **Shell / process execution:** no - **Environment & secrets:** yes - **Dynamic code execution:** no *"Yes" means the capability is present in the source — more access means more to trust, not that it is unsafe.* ## Versions - **0.1.0** — security scan: passed — Imported from the upstream source. ## Links - Listing page: https://agentstack.voostack.com/l/mcp-d0nmega-moltgrid - Seller: https://agentstack.voostack.com/s/d0nmega - Browse the marketplace: https://agentstack.voostack.com/browse --- Listed on AgentStack — the marketplace for AI agent skills and MCP servers. Every listing is security-reviewed. Creators keep 70%.