Skip to main content

Open ACE

Self-hosted control plane for AI coding agents.

Open ACE gives teams a browser workspace for AI coding agents, a remote execution layer for internal machines, and an auditable autonomous workflow system that can take GitHub issues from planning through implementation, review, and merge.

CurrentUpdated from recent PRs
Latest focusAutonomous workflows, run timeline, ZCode support, install hardening
Execution layerRemote Agent, token auth, CLI adapters, terminal relay, app-server mode
Best next readRemote Agent, Deployment, Product Guide, Project status

Core value

From AI sessions to autonomous engineering runs

Issue-driven autonomous development

Turn GitHub issues into planned, executed, reviewed, and merge-ready engineering workflows with milestones, timeline views, and final code-change summaries.

Multi-CLI remote workspace

Run Claude Code, Qwen Code, Codex, OpenClaw, and ZCode on controlled machines with browser terminal, session recovery, and code-server access.

Governance for real team usage

Track API key routing, proxy-token usage, approvals, audit events, quotas, anomalies, compliance, and cost signals.

Why teams adopt it

Designed for teams that need useful automation and real control

For builders

Run real AI development work from issues

Batch GitHub issues, inspect milestone timelines, fork from a previous step, control merge behavior, and keep final code changes visible.

For platform teams

Keep deployment and secrets inside your boundary

Remote Agent, token-based registration, code-server proxying, and API Key Proxy keep machines, credentials, and execution policy self-hosted.

For governance

See what happened while the agent worked

Compact timeline cards, full-text milestone views, run provenance, audit logs, approvals, and usage counters make runs explainable.

Recent momentum

What the current codebase now makes possible

Autonomous workflows became a product surface

Recent work added issue-driven autonomous development, batch navigation, auto-merge controls, retry and timeout handling, fork flows, and session-aware execution.

Execution is easier to inspect

Timeline cards now show milestone summaries, full-text detail, final code changes, status filters, deep links, approvals, and clearer pause or failure semantics.

Remote Agent is broader and safer

ZCode joined Claude Code, Qwen Code, Codex, and OpenClaw, while token management, app-server support, and remote identity checks hardened agent access.

Governance and install paths are getting sharper

Quota enforcement, auditability, API key routing, and Docker, package, macOS, sudoers, migration, and config-path fixes make evaluation and upgrade flows less brittle.

Start here

Evaluate the current product surface