Our first product
Forjal Alpha
A local agent inside the agents you already use.
Four steps. Nothing to set up by hand.
Forjal prepares a model for your machine.
It detects your hardware and sets up a model that fits. No runtimes or quantization to choose.
Your agent delegates.
Claude Code, Codex, OpenClaw, Hermes Agent, OpenCode or Pi hands Forjal a task and chooses which local model runs it.
A local agent does the work.
It reads, runs and writes inside the folders you allow.
Your agent reviews the result.
and keeps going.
What your coding agent can hand off.
Read a codebase.
Search and read across many files and report back what matters.
Run the checks.
Builds, tests and linters, with a summary of what failed.
Dig through logs and documents.
Long logs, transcripts and batches of files.
Draft changes.
Drafts, tests and fixtures, written inside the folders you allow.
You decide what it can touch.
Grant only what the task needs.
Read lets the local agent search and read, and changes nothing.
Write also lets it create or replace files inside the folders you allow.
Run also lets it run commands from Forjal’s allowlist, with no shell. Commands can run your project’s own code, such as its tests.
Forjal gives the local agent no network tools, URLs or credentials. What it reads is not sent to Forjal’s servers. The result goes back to your agent and follows its data policy.
Requirements
| Platform | Requirement |
|---|---|
| macOS | macOS 14 or later on Apple Silicon |
| Windows on Snapdragon | Windows 11 on Arm |
| Windows on Intel or AMD | Windows 11 |
More memory lets Forjal prepare larger models. What each agent can hand off varies by agent.
Your company already owns the compute.
The laptops your team uses carry chips built for AI. Forjal puts them to work inside the agents your team already uses. We are working with a few teams on what that should look like.
Good to know
Does Forjal replace my AI agent?
No. Your agent keeps planning and reviewing. Forjal gives it a local agent to hand work to.
How does my agent connect to Forjal?
Through MCP. The Forjal app connects Claude Code, Codex, OpenClaw, Hermes Agent, OpenCode or Pi, and your agent chooses which local model runs each task.
What goes to the cloud?
What the local agent reads is processed on your machine and is not sent to Forjal’s servers. The result goes back to your agent and follows its data policy.
Does it save tokens?
Local work runs on your hardware, not on metered cloud tokens. How much a whole task saves depends on the work, and measuring it is part of our research.
Which models does it use?
Open models prepared for your hardware. Your agent chooses which one runs.
How do I get Forjal?
Forjal is in early access. Request access and we will send you the download and help you set it up.
How much does it cost?
Forjal is free during early access. For teams, talk to us.
Early access
Forjal is in early access and free while it is. We onboard people in small groups, help them set it up and ship often based on what we hear.