v3.2 · Now with edge inference

Run AI agents at the speed of thought.

NeuralForge is the open runtime framework for building, orchestrating, and scaling AI agents in production — with sub-millisecond tool calls, automatic retries, and built-in observability.

npm install neuralforge

Everything you need to ship agents

A batteries-included runtime, not a pile of glue code.

⚙️

Composable Agent Graphs

Declaratively wire agents, tools, and memory into graphs that branch, loop, and self-correct.

🚀

Sub-ms Tool Dispatch

A zero-copy execution engine routes tool calls with negligible overhead, even at high concurrency.

🔁

Automatic Retries & Fallback

Model timeouts and rate limits are handled for you with configurable backoff and model fallback chains.

📊

Built-in Observability

Every step, token, and tool call is traced out of the box — no extra SDKs to wire up.

🔒

Sandboxed Execution

Run untrusted tool code in isolated sandboxes with fine-grained resource and network policies.

🌐

Model Agnostic

Swap between providers or run locally — one API, any model, no lock-in.

From idea to running agent in minutes

1

Define

Describe your agent's tools, memory, and goals in a few lines of code.

2

Compose

Chain agents into graphs with conditional routing and parallel branches.

3

Deploy

Ship to the NeuralForge runtime and scale from one request to millions.

4

Observe

Trace every run, debug failures, and tune performance from one dashboard.

0 Agents run daily
0 Uptime SLA
0 Median dispatch latency
0 Teams building on NeuralForge

Ready to build your first agent?

Free for individuals and small teams. No credit card required.