Forge builds a property graph of your codebase and uses it to understand structure, track dependencies, detect patterns, and generate context-aware code. No embeddings. No hallucinations about your code.
Get your API keyQuery your codebase as a graph. Find callers, callees, dependencies, and type relationships. Impact analysis shows exactly what breaks when you change a function.
Forge discovers structural patterns in your code through edge-profile clustering. Patterns have a hypothesis lifecycle — proposed, supported, confirmed — with Noisy-OR confidence propagation.
The Cartographer assembles a context envelope for any function: callers, callees, type dependencies, applicable rules, matching patterns. All the context an LLM needs, none of the noise.
The full Forge Cycle: Cartographer assembles context, your LLM generates code, the Auditor checks compilation and rule compliance. Failed attempts get automatic retry with error feedback.
Every node has a temperature. Recently queried or modified code runs hot. Heat diffuses through edges with per-type coefficients. Spreading activation finds relevant context automatically.
Forge is model-agnostic. You bring your Anthropic or OpenAI API key. Forge provides the deterministic graph intelligence — your LLM provides the generation. Any model punches above its weight.
Three deterministic layers. Your LLM only touches one of them.
The graph is deterministic. The reasoning is deterministic. Only the generation step uses an LLM — and even that gets audited.
Free during beta. Get your API key instantly. Bring your own LLM key (Anthropic or OpenAI).
Save this key now. It will not be shown again.