join ushiring 10x engineers →
Works with
Claude CodeCodexOpenCode

The best way to run any AI agent in your company.

One place for your team to run every agent. Easy setup, hard boundaries on what agents can do, and every safeguard open-source.

Join waitlistTalk to team
01Select agent

100+ agents available in the registry

02Configure

Runtime

Integrations

Budget per run

03Run

Select runtime & budget to continue

Every AI agent your company needs. One place to run them all.

01

Run anything. Lock in nothing.

Any agent runtime. Any model. Any tools. One YAML file. When something better ships, swap it in. Your platform never becomes the bottleneck.

No vendor lock-inRuntime-agnostic
02

New agent Friday, running Monday.

Open-source library with community-contributed agents. When a new capability emerges in the ecosystem, your team doesn't file a feature request. They add it.

Community-drivenAlways current
03

Not just engineering.

Co-CEO. Support agent. Executive assistant. This isn't a coding tool with an agent bolted on. It's an operating layer for running AI across your entire company — with every action permissioned and audited.

Whole-companyFull audit trail

Open-source means you’re never waiting on us either.

How it works

One config file. Secrets isolated. Permissions enforced. Every action audited.

01

Agents never see your secrets.

Credentials are injected at runtime through an integration gateway. The model never has access.

Runtime injectionZero exposure
02

Every action is permissioned.

Three modes per action: off, on, or ask — which pauses for human approval.

Granular controlHuman-in-the-loop
03

Config lives in git.

One YAML file defines an agent. Version-controlled. Reproducible. No platform lock-in.

PortableVersion-controlled
oc-agent.yaml
runtime: claude-code
name: pr-reviewer
model: opus
permissions:
  github:
    create_pr: on
    delete_branch: ask  # pauses for human approval
    push_to_main: off
oc-agent.yaml — integrations
integrations:
  github:
    token: vault://github/prod-token
  slack:
    token: vault://slack/bot-token
  linear:
    token: vault://linear/api-key

Open source

You should be able to read the code that controls what AI can do in your company.

Business Source License

Self-host it, fork it, audit it. The commercial layer adds team management and hosted infrastructure — the runtime is yours.

Self-hostableAuditableNo lock-in

Star on GitHub

Follow development, report issues, contribute. The entire agent runtime is public.

useopencompany →