v1.9 — the self-improving agent company framework

Run your agent company on autopilot

Agents that research the internet, implement, verify, and ship their own improvements — with sprint gates, hosted dashboards, API runners, full observability, and humans only where risk demands it.

Open source Python 3.10+ Cursor · Claude · Codex ready
Agent run · API
QA gate passed

Everything you need to run agents like a company

Sprint discipline, autonomous research, self-shipping improvements, and ops — built for teams shipping with AI, not just chatting with it.

Autonomous internet research

Scheduled research missions search and read the web — competitor watches, industry trends, OSS discovery — and file briefs plus scored opportunities into your suggestions queue.

New in v1.9 Web search Missions

Tiered self-improvement

Docs and config auto-apply with an audit log; code lands as auto-PRs that merge only when CI and a review agent approve; schema, auth, and billing always wait for a human. Regression watch auto-reverts bad merges.

New in v1.9 Review agent Auto-revert

Self-improving intelligence

An intelligence agent synthesizes every performance signal into evidence packs and digests, scores whether past suggestions actually worked, and vets new OSS libraries before they're adopted.

Evidence packs Suggestion scoring OSS vetting

Sprint board & phase gates

Plan → execute → QA → review with agent roles, owner assignments, and enforced gates so nothing ships untested.

Kanban Manual QA Review sign-off

API agent runner

Run chassis agents and flows via Anthropic, OpenAI, Cursor Cloud Agents, or local Ollama — with tool loops, web search and fetch, branch picker, and run transcripts.

Anthropic OpenAI Cursor Ollama

Hosted dashboards

Private subdomains with TLS, dual login (GitHub OAuth + password), and 24/7 state mirrors synced from your repo.

Multi-tenant OAuth Sync-back

Scheduler & triggers

Cron schedules plus inbound webhooks from GitHub, Slack, and Teams — fan-out to Slack, Teams, and Linear when runs finish.

Webhooks Cron Notifications

Observability & run logs

Audit trail for every agent run, budget caps, daily limits, and a dashboard Run Logs tab with hook forwarding from dev machines.

Budget Audit trail Linear run-log

Agent company chassis & autopilot

Developer, QA, research, sales, and custom roles with skills, flows, and learned memory — plus a full-autonomy orchestration loop governed by a kill switch, budget caps, and daily run limits.

Roles Autopilot Kill switch
GitHub Linear Slack Teams PyPI Cursor

From zero to agent company in three steps

Adopt in any repo — consumer apps, internal tools, or the framework itself.

1

Install & init

One pip install, then company-os init compiles agents, rules, and sprint state into your host IDE.

2

Run sprints & agents

Work tasks through phase gates locally or trigger flows via API runner, scheduler, and webhooks.

3

Host & observe

Push state to a private dashboard subdomain — your team signs in, reviews runs, and stays in sync.

Teams running on company-os

Hosted consumer dashboards with secure sign-in. Loading…

Loading teams…

Ready to run a company that improves itself?

Install from PyPI, adopt in five commands, or explore the docs to see how production teams use company-os today.