Skip to main content

What Open Forge doesn't do

Some limits to know before you adopt it:

  • It doesn't make a model deterministic. Open Forge makes the right context cheap to find and the rules explicit. An agent can still misread or skip them, so review stays with you.
  • It depends on your harness reading AGENTS.md (or CLAUDE.md through the bridge). Everything starts there.
  • It doesn't set up your harness. Hooks, custom agents, and your harness's own instruction files stay yours to configure. Open Forge is built to sit beside them, as the introduction explains.
  • It isn't a methodology. There's no mandatory workflow, role, or ceremony, and the base Framework ships no workflows. Extensions offer methods, but installing one doesn't make its methods required.
  • It doesn't record everything. Memory holds what's worth keeping, and the rules ask agents to save outcomes deliberately rather than log every conversation.
  • It isn't a runtime. There's no hidden database, background process, or agent host. Tags and routes are conventions that agents follow by reading them, not code that enforces them. A LoadNow tag tells an agent to read a file. It can't prove the agent did.
  • It doesn't grant permissions. Installing a workflow or finishing one never authorizes commits, merges, publishing, or contacting external systems. Your harness's permission model still applies.

The CLI is still in beta​

The CLI is in public beta. Command output and some behaviors may still change before 1.0. The Framework files themselves work with or without the CLI.

Found something rough? Open an issue.

Next: Try Open Forge in the ten-minute guide.