Forge
Generate text embeddings and list Forge models from your agent when building RAG, search, or semantic features without custom HTTP glue.
Overview
Forge MCP is an MCP server for the build phase that calls Voxell Forge’s hosted embedding API via embed and list_models tools over stdio.
What is this MCP server?
- Two documented tools: embed and list_models against Voxell Forge
- Stdio npm package @voxell/forge-mcp (version 0.1.5)
- Requires FORGE_API_KEY from https://dash.voxell.ai with free starter tokens
- Open-source forge-mcp repository on GitHub (VoxellInc)
- 2 MCP tools documented: embed and list_models
- Package and server metadata version 0.1.5
What problem does it solve?
You are wiring RAG or semantic search in an agent session and keep rewriting embedding HTTP requests instead of stable tools.
Who is it for?
Indie builders using Voxell Forge who want stdio MCP embedding tools inside Claude Code or Cursor.
Skip if: Teams that need local-only embeddings, multi-modal vectors without Forge, or a full vector database MCP.
What do I get? / Deliverables
Your agent lists Forge models and returns embeddings through MCP so you can focus on indexing and retrieval logic.
- Embedding vectors from Forge via the embed tool
- Available model list from list_models for pipeline choices
Recommended MCP Servers
Journey fit
How it compares
Focused embedding API MCP (2 tools), not a complete RAG stack or generic LLM chat server.
Common Questions / FAQ
Who is Forge MCP for?
Developers building agent or SaaS features that need Voxell Forge text embeddings exposed as MCP tools.
When should I use Forge MCP?
During build when you prototype or ship semantic search, clustering, or RAG and want list_models plus embed from the agent.
How do I add Forge MCP to my agent?
Install @voxell/forge-mcp from npm, set FORGE_API_KEY in the MCP server environment, and register the stdio server in your client’s mcp.json per Voxell’s repo instructions.