Runjam

RunJam is a local first desktop manager for AI coding agents, covering Claude Code, Codex CLI and Gemini CLI. It brings parallel agent sessions into one window with a sidebar and per session workspaces, so developers no longer juggle several terminal tabs, config files and API keys. It scans the system path for the supported agents and can install, update and configure them in one click. A unified model hub lets a user define a model once and assign it to each agent, including DeepSeek, Kimi and Qwen models, and a built in protocol proxy translates between Anthropic, OpenAI and Gemini formats on the fly so agents can run models from other vendors. To manage spend, it detects prompt caching, keeps a local response cache and shows a per session cost dashboard. Sessions persist with full text search and an archive, and the app is described as agent agnostic, requiring no modification of the agents themselves. Data stays on the user's machine in a local directory, API keys are stored in the operating system keychain, and telemetry is off by default. The source is available on GitHub. RunJam is aimed at developers who use several AI coding agents every day and want one place to switch models, track costs and keep history. It runs on macOS, Windows and Linux, is released under the MIT license and is a free download. The site lists no paid plans.