
1 / 3
【Overview】 jcode is an open-source AI coding agent that runs in the terminal, written in Rust and designed for developers who push many agents at once. It targets demanding development workflows where a single agent is not enough: jcode lets you run many coding agents in parallel with low memory overhead per additional session, so a single machine can carry a fleet of workers. The product is built around the reality that agentic coding is a compute problem as much as a tooling problem. The agent itself is free and works with your own models, while hosted inference, cloud sessions, and remote Rust builds are available as paid add-ons. A self dev mode lets the agent edit its own source, build, test, and reload the binary and keep going, which makes the harness itself part of the workflow. jcode is aimed at developers running many agents in parallel, at demanding development workflows, and at teams and enterprises that want one local or remote control plane for long-running sessions. It is not intended for executing untrusted code without a container or VM, since the SDK's isolation is not a sandbox. 【Key features】 • Multi-agent swarms: Runs many coding agents in parallel, with low memory overhead for each additional session. • Self dev mode: Lets the agent modify its own source code, build, test, reload its binary, and continue automatically. • Hill-climbability rating: Scores every agent goal from 0 to 100 for how quantifiable and iterable its progress is, and pushes back on weak goals. • Auto-poke persistence: Detects unfinished todos and prompts the agent to keep going, including in headless `jcode run` sessions. • MCP servers and skills: Loads MCP servers and markdown skill packs, injecting skills when the conversation matches. • Remote daemon and TypeScript SDK: Runs the daemon over SSH with socket forwarding, exposes a WebSocket gateway on port 7643 for paired phone and web clients, and lets TypeScript apps launch or attach to sessions with streamed replies, tool calls, and permission prompts. 【Use cases】 1. Run 20 coding agents in parallel on one machine. 2. Resume long-running agent tasks after disconnects. 3. Automate coding sessions from TypeScript apps. 【Pricing】 Open source agent: $0 (MIT licensed, open source, Bring your own models) Token plan: $10/month — $20 monthly inference credit (double the price), Jcode Cloud with 50 machine hours per month, access to Claude Opus 5.5, Sonnet 5.5, Haiku 5.5, GPT-6.1 Sol, GPT-6 Sol, Astra, Luna; remote Rust builds not included Token plan $20: $20/month — $40 monthly inference credit, 1,000 Jcode Cloud machine hours per month, 5,000 remote Rust builds per month, first month free with 5,000 builds 【Pros and cons】 Pros: - Developers who already run several agents side by side get the most value: the swarm model and auto-poke persistence reduce babysitting, so long workflows keep moving without constant re-prompting. - Self dev mode and the hill-climbability rating suit teams that want measurable, iterable goal definitions rather than open-ended prompts that stall. - The free MIT-licensed agent with bring-your-own-models is an unusually low-friction entry point for individual developers, while the paid tiers scale compute for teams and enterprises. Cons: - Only stdio MCP servers are supported; HTTP and SSE entries are skipped. - The WebSocket gateway is off by default and listens on port 7643. - The TypeScript SDK requires Node 20 or newer, and the Windows SDK has no live end-to-end coverage yet. - The desktop app is in beta, and the Windows app is not code-signed yet. - The paid $10/month token plan does not include remote Rust builds; those start on the $20/month plan. 【FAQ】 Q: Is jcode free to use? A: Yes. The agent itself is open source under the MIT license at $0, and you can bring your own models. Hosted inference, Jcode Cloud sessions, and remote Rust builds are optional paid plans starting at $10/month. Q: Who is jcode designed for? A: It is built for developers running many agents in parallel and for demanding development workflows, including teams and enterprises. It is not meant for executing untrusted code without a container or VM. Q: How does the $20/month token plan differ from the $10/month plan? A: The $20/month plan includes $40 monthly inference credit, 1,000 Jcode Cloud machine hours per month, and 5,000 remote Rust builds per month, plus a first month free with 5,000 builds. The $10/month plan offers $20 monthly inference credit and 50 Jcode Cloud machine hours per month, without remote Rust builds.
Open source agent · $0 (MIT licensed, open source, Bring your own models) Token plan · $10/month ($20 monthly inference credit (double the price), Jcode Cloud with 50 machine hours per month, Access to Claude Opus 5.5, Sonnet 5.5, Haiku 5.5, GPT-6.1 Sol, GPT-6 Sol, Astra, Luna) Token plan $20 · $20/month ($40 monthly inference credit, 1,000 Jcode Cloud machine hours per month, 5,000 remote Rust builds per month)
WPS AI is an AI assistant built into WPS Office that helps users write documents, generate presentation outlines and…
Baidu Comate is an AI coding assistant built by Baidu on its ERNIE model. It goes beyond simple autocomplete: it…
UniConverter is an AI media conversion and enhancement tool for video, audio, and images. It combines everyday…
6 similar tools to consider instead of jcode