{"slug": "prime-agent-a-self-improving-rlm-agent", "title": "Prime Agent: A Self-Improving RLM Agent", "summary": "Prime Intellect released Prime Agent, an open-source coding and research agent designed for long-running tasks, featuring a persistent IPython environment, recursive subagents, and a self-improving harness that refines supplemental state without altering the base system prompt. The agent supports background sessions, direct inter-agent communication, and executable skills, and is installable on macOS and Linux via a curl command.", "body_md": "[Documentation](/PrimeIntellect-ai/prime-agent/blob/main/packages/coding-agent/docs/index.md) •\n[Verifiers](https://github.com/PrimeIntellect-ai/verifiers) •\n[PRIME-RL](https://github.com/PrimeIntellect-ai/prime-rl) •\n[pi-mono](https://github.com/badlogic/pi-mono)\n\nPrime Agent is an open-source coding and research agent for general and long-running work. It is designed around two core abstractions:\n\n- The\ntreats context as variables ([Recursive Language Model (RLM)](https://www.primeintellect.ai/blog/rlm)*prompt-as-a-variable*) and tools like recursive subagents as function calls (*programmatic tool /sub-agent calling*) inside a persistent REPL. - The\nstores supplemental prompts, memories, skill descriptions, and reusable subagent specifications as durable state that Prime Agent can refine through small, evidence-backed updates, local to the session by default.[Continual Harness](https://arxiv.org/abs/2605.09998)\n\nPrime Agent combines a persistent Python control environment with durable harness state, so useful working context and reusable operating patterns can outlive a single chat window.\n\n**Everything is programmatic:** persistent IPython is the built-in model tool; file operations, shell commands, tool use, subagents, and context management happen through code.**Subagents are built in:**`rlm(...)`\n\nspawns real child agents for parallel or background work and returns their results programmatically.**The harness can improve:**`/refine`\n\nreviews the current trajectory and can apply small, evidence-backed updates to supplemental harness state. It never rewrites the immutable base system prompt, and recorded snapshots support rollback.**Skills are executable:** skills are importable Python packages, and the built-in skill creator can turn recurring workflows into project or personal skills.**Sessions run in the background:** daemon-backed agents keep running when the terminal disconnects and can be reattached later.**Agents communicate directly:** running agents can exchange messages and orchestrate one another without routing everything through the user.**Long tasks keep moving:** automatic compaction, persistent goals, heartbeats, schedules, autonomous mode, and retained subagents preserve progress across turns and terminal sessions.\n\nInstall the latest stable release on macOS or Linux:\n\n```\ncurl -fsSL https://app.primeintellect.ai/prime-agent/install.sh | sh\n```\n\nThe installer downloads a versioned release, verifies its SHA-256 checksum, installs the `prime-agent`\n\ncommand, and can prepare the IPython runtime used by the agent.\n\nStart Prime Agent from the repository or directory you want it to work in:\n\n```\ncd /path/to/project\nprime-agent\n```\n\nOn first launch, run `/login`\n\nto choose a subscription or API-key provider. Prime Agent works in the current directory and can run commands and modify files there. Use a disposable clone, clean worktree, or another checkpoint you can inspect and restore.\n\nWarning\n\nPrime Agent executes model-generated Python and project commands with your user permissions. Its worker and kernel processes improve lifecycle isolation and recovery; they are **not** a security sandbox. Review changes and use trusted repositories, instructions, skills, and extensions only. Run untrusted code or instructions in an external sandbox or restricted environment.\n\nUseful commands:\n\n```\nprime-agent agents                   # Browse running, idle, and saved sessions\nprime-agent attach <agent>           # Reattach to a running session\nprime-agent --resume <path|id>       # Resume a saved session\nprime-agent status                   # Inspect background service state\nprime-agent doctor [--fix]           # Inspect or repair background services\nprime-agent update [--force]         # Update Prime Agent\nprime-agent shutdown [--force]       # Stop every agent, worker, and background service\n```\n\nPrime Agent is built for long-running work, especially for evaluations in research. These features are available in the TUI, and when run autonomously.\n\n**Continual Harness:**`/refine`\n\ncan persist focused, reviewable lessons as supplemental prompts, memories, reusable skill descriptions, or subagent specifications, with recorded refinement history. It does not replace packaging and reviewing new executable skills.**Direct agent-to-agent communication:** running agents and retained subagents can discover one another, exchange messages, and steer active work.**Daemon-backed continuity:** active sessions, IPython state, schedules, and subagents keep running when the terminal detaches and can be reattached later.**Heartbeats and schedules:**`/heartbeat`\n\n,`rlm_heartbeat`\n\n, and`prime-agent schedule`\n\ncan re-enter a session periodically or at a specific time.**Persistent goals:**`/goal`\n\nkeeps an objective and its progress active across turns until it is completed, paused, or cleared.**Bounded autonomous mode:**`/autonomous`\n\ncontinues within configured turn, token, and time budgets and can run user-defined quality gates. A passed gate checks only what that gate verifies; reaching a limit does not imply task success.\n\n[Quickstart](/PrimeIntellect-ai/prime-agent/blob/main/packages/coding-agent/docs/quickstart.md)— install, authenticate, and run a first session[Usage and CLI reference](/PrimeIntellect-ai/prime-agent/blob/main/packages/coding-agent/docs/usage.md)— commands, sessions, autonomous limits, and output modes[Long-running and background agents](/PrimeIntellect-ai/prime-agent/blob/main/packages/coding-agent/docs/long-running-agents.md)— detach and reattach, goals, heartbeats, and schedules[RLM programming model](/PrimeIntellect-ai/prime-agent/blob/main/packages/coding-agent/docs/rlm.md)— persistent IPython, subagents, skills, and the trust model[JSON mode](/PrimeIntellect-ai/prime-agent/blob/main/packages/coding-agent/docs/json.md)and[RPC mode](/PrimeIntellect-ai/prime-agent/blob/main/packages/coding-agent/docs/rpc.md)— headless automation and integrations[Skills](/PrimeIntellect-ai/prime-agent/blob/main/packages/coding-agent/docs/skills.md)— install and create reusable capabilities[Provider setup](/PrimeIntellect-ai/prime-agent/blob/main/packages/coding-agent/docs/providers.md)— subscription and API-key providers[Architecture overview](/PrimeIntellect-ai/prime-agent/blob/main/packages/coding-agent/docs/architecture.md)— daemon, worker, kernel, and persistence boundaries[Development](/PrimeIntellect-ai/prime-agent/blob/main/packages/coding-agent/docs/development.md)— build and run from source\n\nOur agent and TUI is built on top of [ pi](https://github.com/earendil-works/pi). We thank the authors of\n\n`pi`\n\nfor their valuable work.Prime Agent is fully open source and released under the [MIT License](/PrimeIntellect-ai/prime-agent/blob/main/LICENSE).", "url": "https://wpnews.pro/news/prime-agent-a-self-improving-rlm-agent", "canonical_source": "https://github.com/PrimeIntellect-ai/prime-agent", "published_at": "2026-08-10 08:27:22+00:00", "updated_at": "2026-08-10 08:41:00.840434+00:00", "lang": "en", "topics": ["artificial-intelligence", "ai-agents", "ai-tools", "ai-research"], "entities": ["Prime Intellect", "Prime Agent", "IPython", "PRIME-RL", "pi-mono"], "alternates": {"html": "https://wpnews.pro/news/prime-agent-a-self-improving-rlm-agent", "markdown": "https://wpnews.pro/news/prime-agent-a-self-improving-rlm-agent.md", "text": "https://wpnews.pro/news/prime-agent-a-self-improving-rlm-agent.txt", "jsonld": "https://wpnews.pro/news/prime-agent-a-self-improving-rlm-agent.jsonld"}}