Open source · MIT · building in public

The whole platform.
Open source. All of it.

No closed core. No enterprise gate. The knowledge pipeline, the tool registry, the agent runtime, the chatbot widget — every line is on GitHub under MIT. Use the hosted product for convenience, or run it yourself. Your choice, always.

See the packages GitHub repo at launch
MIT
License
TypeScript
Primary language
pre-launch
Status
public
Roadmap
Inside the repo

Packages — each one earns its place.

Each package is independently useful. Pull one into your own stack, or use them all together through the hosted platform.

agent/knowledge
planned

Ingestion pipeline. Chunk documents, generate embeddings, and store them in a vector index ready for semantic retrieval.

TypeScript · MIT
agent/connectors
planned

Data source adapters — Amazon S3, SharePoint, and local file upload. Syncs on a schedule and keeps the index fresh.

TypeScript · MIT
agent/tools
planned

Tool builder and registry. Define any REST API as an agent tool with a name, description, and parameter schema.

TypeScript · MIT
agent/orchestrator
planned

Agent runtime. Routes queries through the knowledge layer and tool calls, manages multi-agent handoffs, enforces step limits.

TypeScript · MIT
agent/mcp
planned

MCP server support. Plug any MCP-compatible tool or server directly into an agent — no adapter code needed.

TypeScript · MIT
agent/widget
planned

Embeddable chatbot UI. One script tag, streaming responses, source citations, and tool-call indicators out of the box.

TypeScript · MIT
Why open source

The commitments we're making.

Open source isn't a marketing move. It's a constraint we're choosing so you never have to trust us blindly.

→ No closed core

Everything that ships, ships under MIT. There is no "enterprise edition" with the good parts locked away. What you see on GitHub is what powers the product.

→ Transparent build

Commits, issues, and roadmap decisions happen in public. If we change direction, you can read why in the issue thread, not in a changelog months later.

→ Replaceable parts

Each package is independently useful. You should be able to take the knowledge ingestion layer and use it without the rest. No forced coupling to the hosted platform.

→ Your data, your repo

Self-hosting is a first-class path, not an afterthought. The hosted product is convenience — the open-source repo is the escape hatch if you ever need it.

Contribute

Build this with us.

Forge will be built in the open, commit by commit. The repo goes public at launch — when it does, the most useful thing you can do won't be submitting PRs. It'll be using it on a real problem and telling us what breaks.

01
Star the repo
Helps others find it. Takes two seconds.
02
Open an issue
Found a bug, missing connector, or confusing behaviour? Tell us. Honest feedback shapes the roadmap directly.
03
Build something
Use it. Break it. Share what you built. That feedback is worth more than a PR to us at this stage.
Follow the build → github.com/cuecoder/forge · at launch
Build roadmap
Knowledge ingestion pipeline
packages/knowledge · planned
S3 connector + auto-chunking
packages/connectors · planned
Tool builder — REST API schema
packages/tools · planned
Agent orchestrator + handoffs
packages/orchestrator · planned
Embeddable chat widget
packages/widget · planned
Repo goes public at launch — follow along soon.