Open-source self-improving agent harness that takes outcome-based tasks like reporting, email cleanup and integrations, running in a terminal with any LLM.
darrenhinde/OpenAgentsControl
An agent framework for OpenCode that adds plan-first, approval-gated AI coding workflows using your team's own coding patterns.
- Stars
- 4,938
- Forks
- 402
- Open issues
- 70
- Last commit
- 6 days ago
What is it?
OpenAgents Control (OAC) is a set of AI agents built on top of the OpenCode CLI. Its agents load context files that describe your stack, naming conventions and security rules before writing code, so generated output follows the project's existing patterns rather than generic defaults.
The workflow is plan-first: an agent analyzes a request, proposes an implementation plan, waits for approval, then executes step by step with validation such as type checks, tests and code review delegated to specialist subagents. Agents are plain markdown files that can be edited directly, and the context files can be committed to a repository so a team shares the same standards.
OAC works with any model OpenCode supports and includes a Claude Code plugin in beta. The project is currently targeting OpenCode v1, and the README notes that a larger redesign is planned.
Key features
- 01Plan-first workflow where the agent proposes steps and waits for approval before editing files or running commands.
- 02Context files describing your stack and conventions are loaded before code generation.
- 03Specialist subagents handle tasks such as testing, code review and build validation.
- 04Agents are editable markdown files, with per-agent model configuration.
- 05Works with any model provider supported by OpenCode, plus a beta Claude Code plugin.
Use cases
- Adding features to an existing codebase while keeping code consistent with established patterns.
- Sharing one set of coding standards across a team by committing context files to the repository.
- Working on production code where you want to review a plan before any changes are made.
Quick start
- 1
Install via the one-command installer from the README
- 2
Start OpenCode with the OpenAgent agent
opencode --agent OpenAgent - 3
Ask the agent to build a feature, then approve its plan
Things to know
- Targets OpenCode v1, and a major upcoming release will replace much of the current system.
- Installation relies on a curl-piped shell script, and the Claude Code plugin is still in beta.
In collections
Topics
FAQ
- What is OpenAgentsControl?
- An agent framework for OpenCode that adds plan-first, approval-gated AI coding workflows using your team's own coding patterns. OpenAgents Control (OAC) is a set of AI agents built on top of the OpenCode CLI. Its agents load context files that describe your stack, naming conventions and security rules before writing code, so generated output follows the project's existing patterns rather than generic defaults.
- Is OpenAgentsControl free to use? What license does it have?
- OpenAgentsControl is open source under the MIT license. Check the license terms before using it in commercial products.
- Is OpenAgentsControl still maintained?
- Actively maintained. The last commit to OpenAgentsControl was on Oct 19, 2026.
- What can I use OpenAgentsControl for?
- Adding features to an existing codebase while keeping code consistent with established patterns. Sharing one set of coding standards across a team by committing context files to the repository. Working on production code where you want to review a plan before any changes are made.
- How do I install OpenAgentsControl?
- Install via the one-command installer from the README Start OpenCode with the OpenAgent agent: opencode --agent OpenAgent Ask the agent to build a feature, then approve its plan
- Where can I get OpenAgentsControl?
- The source code and installation instructions are on GitHub: https://github.com/darrenhinde/OpenAgentsControl
Summary generated by AI from the project's README. GitHub stats refresh daily.
Related projects
A curated directory of 1000+ Claude Skills and plugins for Claude Code, Claude.ai and coding agents like Codex, Cursor and Gemini CLI.
Open-source workflow automation platform and alternative to Zapier, with TypeScript-based integrations that also run as MCP servers for AI agents.
A curated list of workflow automation software, AI agent platforms, frameworks, and learning resources for developers and teams.
A framework for autonomous Android and computer use, letting a planner and finder LLM operate phones and desktops from natural-language tasks.
A prompt-native development system where .prompt files are the source and Python, TypeScript, Go and other code is generated from them.