Prime Agent: an open source coding and research agent for long-running work

Prime Agent (Prime Intellect)

The agent that writes code should not be the one that judges it, and Prime Agent is built around that rule. It is Prime Intellect's open source coding and research agent for long running work, now rebuilt in Rust. Its core idea is the Recursive Language Model: context is treated as variables and subagents are called like functions inside a persistent Python REPL. A Continual Harness then stores memories, skills, prompts, and subagent specs as durable state that the agent can refine in small, evidence backed updates. The stable release is v0.10.2 from 28 October 2026, and the repo now has 22,632 stars. Prime Intellect says more than 2,000 agents carried out the Rust rewrite.

What you get

• A reviewer separate from the writer, so the agent that writes the code is not the one judging it
• Durable state across sessions: memories, skills, prompts and subagent specs persist in the Continual Harness
• Subagents called like functions, with context treated as variables inside a persistent Python REPL
• A Rust rebuild for long running work, at stable release v0.10.2 and MIT licensed per the README

Try it on a long running task first, and keep the review step separate from the writing step.

Repo: https://github.com/PrimeIntellect-ai/prime-agent

Quick poll

What would you use Prime Agent (Prime Intellect) for first?

• Long-running coding work
• Research tasks
• The subagent-as-function approach
• Watching it for now

Vote in the comments.

---

If you found Prime Agent (Prime Intellect) interesting, I write about tools like this every week in The AI-Native Engineering Leader, what actually works for engineering leaders building with AI agents, and what to skip.