runs anywhere. uses anything
An open-source coding agent for the CLI. Cloud APIs, gateways, and local models — same tools, same agents, same workflow.
Install · Quick start · Providers · Sessions · Config · Docs
Built from source. Windows examples below; macOS and Linux are identical minus the shell syntax.
Prerequisites — Node =22 (enforced by engines.node) and Bun.
No Bun? winget install Oven-sh.Bun or bun.sh.
That's it — openclaude is now on your PATH.
Run /provider inside OpenClaude for guided setup with saved profiles — this is the recommended path. Credentials land in .openclaude-profile.json.
OpenClaude talks to Ollama's native chat API and requests a 32768-token window per request, so same-session history isn't silently trimmed by the OpenAI-compat shim. Override with OPENCLAUDEOLLAMANUMCTX or OLLAMACONTEXTLENGTH.
Run /onboard-github inside OpenClaude. Interactive, credentials saved.
Project .env files are not auto-loaded. Use --provider-env-file .env for provider vars, or export runtime knobs from your shell.
Provider Setup Key detail --------- OpenAI-compatible /provider · env Any /v1 server — OpenRouter, DeepSeek, Groq, Mistral, LM Studio Ollama /provider · env Local, no API key Gemini /provider · env API key only GitHub Models /onboard-github Saved credentials Codex / Codex OAuth /provider Browser sign-in or existing Codex CLI auth Gitlawb Opengateway default · /provider Startup default on fresh installs; get a key Bedrock · Vertex · Foundry env Anthropic-family routes
From the project README.
Add the radar badge to your README — it shows your project was picked up by MCP Radar and links to this page:
[](https://mcp.liqiwa.com/s/0xwilliamortiz--openclaude-improved.html)
Cẩm nang thực chiến Mastering Claude Code (tiếng Việt): 16 chương biên soạn từ tài liệu chính thức của Anthropic — CLI, workflows, MCP, subagents, skills, hooks, plugins, CI/CD & tối ưu context.
12122J/mcpvetMCP security scanner — vet a Model Context Protocol server before you add it to Claude Code, Cursor, or Windsurf. Grades it A–F, catching credential theft, tool-poisoning, and install-script payloads
Gowthaman90/mcp-bastionReliability + security proxy for the Model Context Protocol (MCP): self-healing connections, runtime tool-security (rug-pull/poisoning detection), and a compliance-mapped audit trail.
ZoeLinUTS/MagicTeX-mcpMagicTeX — a LaTeX editor for AI agents (Claude Code): live PDF preview, PDF comments that drive edits, Visual (WYSIWYG) mode, and git-tracked history. No local TeX install.
dhavanikgithub/chartbrew-mcpChartbrew + AI agents. MCP server exposing Chartbrew's documented API: teams, connections, datasets, dashboards, charts, live queries, and secure embedding. TypeScript · stdio · restricted/unrestricte
hoangann2000/figma-mcp-consoleMCP server that lets your AI assistant read and write Figma in real time — no API token, no rate limits. Works with Claude, Cursor, Copilot, Codex and any MCP client.
The top new MCP servers of the week, every Monday. No spam, unsubscribe anytime.