CrewCoder: An Open-Source, Security-Conscious Coding Agent Platform

CrewCoder is a new open-source coding-agent platform that positions itself as an extensible and security-conscious alternative in the increasingly crowded field of AI developer tools. Built as an npm-workspaces monorepo, it requires Node.js 22 or newer and offers a terminal UI (TUI) alongside a command-line interface. The umbrella package, crewcoder, includes both the agent and the TUI, and can be run via npx without a persistent global install, or installed globally to expose both commands.

At its core, CrewCoder provides a harness that integrates with a wide range of AI models and coding agents, including Claude Code, Codex, OpenCode, Grok, OpenAI, Anthropic, OpenRouter, xAI, DeepSeek, and Mistral. It supports a variety of transport mechanisms (process, HTTP/SSE, WebSocket) and features first-class support for reasoning streams. The platform is designed for interactive development, multi-worker orchestration, durable sessions, detached goals, and remote execution, making it a versatile tool for individual developers and teams.

One of the standout features is its extensive set of local tools: read/write/edit with transactional edits, grep, list-files, bash and background jobs, git primitives, and LSP-backed code intelligence. It also includes memory, worker delegation, and plugin scaffolding. Security is a priority, with multiple approval modes (never, review, always, full-access, sandboxed), sandbox and trust tiers, runtime guardrails, audit logging, and redaction. The fleet API requires an automatically generated private bearer token for remote execution.

CrewCoder is highly extensible via capability-based extensions declared in crewcoder.extension.json, covering providers, tools, skills, prompt packs, commands, workflows, context providers, validators, approval policies, hooks, and sandboxed Live UI components. The project also offers an SDK for custom interfaces and a client for browser-based usage. A CI integration is provided via a GitHub composite action, and GitLab support is included. The authors note that the extension registry is still in progress, suggesting ongoing development.

The platform's focus on security, extensibility, and multi-agent orchestration makes it an interesting option for developers who want more control over their AI-assisted workflows. While the ecosystem is young, the breadth of features and attention to safety could appeal to teams looking for a self-hosted, customizable solution.

For more details, see the CrewCoder documentation.