[{"data":1,"prerenderedAt":4},["ShallowReactive",2],{"readme:prime-agent":3},"\u003Cp align=\"center\">\n  \u003Ca href=\"https:\u002F\u002Fprimeintellect.ai\" rel=\"nofollow ugc noopener\">\n    \u003Cpicture>\n      \u003Csource media=\"(prefers-color-scheme: light)\" srcset=\"https:\u002F\u002Fgithub.com\u002Fuser-attachments\u002Fassets\u002F40c36e38-c5bd-4c5a-9cb3-f7b902cd155d\">\u003C\u002Fsource>\n      \u003Csource media=\"(prefers-color-scheme: dark)\" srcset=\"https:\u002F\u002Fgithub.com\u002Fuser-attachments\u002Fassets\u002F6414bc9b-126b-41ca-9307-9e982430cde8\">\u003C\u002Fsource>\n      \u003Cimg alt=\"Prime Intellect\" src=\"https:\u002F\u002Fgithub.com\u002Fuser-attachments\u002Fassets\u002F6414bc9b-126b-41ca-9307-9e982430cde8\" width=\"312\" \u002F>\n    \u003C\u002Fpicture>\n  \u003C\u002Fa>\n\u003C\u002Fp>\u003Ch3>\nPrime Agent: A Self-Improving RLM Agent\n\u003C\u002Fh3>\u003Cp align=\"center\">\n  \u003Ca href=\"https:\u002F\u002Fgithub.com\u002FPrimeIntellect-ai\u002Fprime-agent\u002Fblob\u002FHEAD\u002Fpackages\u002Fcoding-agent\u002Fdocs\u002Findex.md\" rel=\"nofollow ugc noopener\">Documentation\u003C\u002Fa> •\n  \u003Ca href=\"https:\u002F\u002Fgithub.com\u002FPrimeIntellect-ai\u002Fverifiers\" rel=\"nofollow ugc noopener\">Verifiers\u003C\u002Fa> •\n  \u003Ca href=\"https:\u002F\u002Fgithub.com\u002FPrimeIntellect-ai\u002Fprime-rl\" rel=\"nofollow ugc noopener\">PRIME-RL\u003C\u002Fa> •\n  \u003Ca href=\"https:\u002F\u002Fgithub.com\u002Fbadlogic\u002Fpi-mono\" rel=\"nofollow ugc noopener\">pi-mono\u003C\u002Fa>\n\u003C\u002Fp>\u003Cp align=\"center\">\n  \u003Ca href=\"https:\u002F\u002Fgithub.com\u002FPrimeIntellect-ai\u002Fprime-agent\u002Factions\u002Fworkflows\u002Fci.yml\" rel=\"nofollow ugc noopener\">\n    \u003Cimg src=\"https:\u002F\u002Fgithub.com\u002FPrimeIntellect-ai\u002Fprime-agent\u002Factions\u002Fworkflows\u002Fci.yml\u002Fbadge.svg\" alt=\"CI\" \u002F>\n  \u003C\u002Fa>\n  \u003Ca href=\"https:\u002F\u002Fgithub.com\u002FPrimeIntellect-ai\u002Fprime-agent\u002Factions\u002Fworkflows\u002Fbuild-binaries.yml\" rel=\"nofollow ugc noopener\">\n    \u003Cimg src=\"https:\u002F\u002Fgithub.com\u002FPrimeIntellect-ai\u002Fprime-agent\u002Factions\u002Fworkflows\u002Fbuild-binaries.yml\u002Fbadge.svg\" alt=\"Build Binaries\" \u002F>\n  \u003C\u002Fa>\n\u003C\u002Fp>\u003Cp>Prime Agent is an open-source coding and research agent for general and long-running work. It is designed around two core abstractions:\u003C\u002Fp>\n\u003Cul>\n\u003Cli>The \u003Cstrong>\u003Ca href=\"https:\u002F\u002Fwww.primeintellect.ai\u002Fblog\u002Frlm\" rel=\"nofollow ugc noopener\">Recursive Language Model (RLM)\u003C\u002Fa>\u003C\u002Fstrong> treats context as variables (\u003Cem>prompt-as-a-variable\u003C\u002Fem>) and tools like recursive subagents as function calls (\u003Cem>programmatic tool \u002Fsub-agent calling\u003C\u002Fem>) inside a persistent REPL.\u003C\u002Fli>\n\u003Cli>The \u003Cstrong>\u003Ca href=\"https:\u002F\u002Farxiv.org\u002Fabs\u002F2605.09998\" rel=\"nofollow ugc noopener\">Continual Harness\u003C\u002Fa>\u003C\u002Fstrong> stores 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.\u003C\u002Fli>\n\u003C\u002Ful>\n\u003Cp>Prime 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.\u003C\u002Fp>\n\u003Cul>\n\u003Cli>\u003Cstrong>Everything is programmatic:\u003C\u002Fstrong> persistent IPython is the built-in model tool; file operations, shell commands, tool use, subagents, and context management happen through code.\u003C\u002Fli>\n\u003Cli>\u003Cstrong>Subagents are built in:\u003C\u002Fstrong> \u003Ccode>rlm(...)\u003C\u002Fcode> spawns real child agents for parallel or background work and returns their results programmatically.\u003C\u002Fli>\n\u003Cli>\u003Cstrong>The harness can improve:\u003C\u002Fstrong> \u003Ccode>\u002Frefine\u003C\u002Fcode> reviews 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.\u003C\u002Fli>\n\u003Cli>\u003Cstrong>Skills are executable:\u003C\u002Fstrong> skills are importable Python packages, and the built-in skill creator can turn recurring workflows into project or personal skills.\u003C\u002Fli>\n\u003Cli>\u003Cstrong>Sessions run in the background:\u003C\u002Fstrong> daemon-backed agents keep running when the terminal disconnects and can be reattached later.\u003C\u002Fli>\n\u003Cli>\u003Cstrong>Agents communicate directly:\u003C\u002Fstrong> running agents can exchange messages and orchestrate one another without routing everything through the user.\u003C\u002Fli>\n\u003Cli>\u003Cstrong>Long tasks keep moving:\u003C\u002Fstrong> automatic compaction, persistent goals, heartbeats, schedules, autonomous mode, and retained subagents preserve progress across turns and terminal sessions.\u003C\u002Fli>\n\u003C\u002Ful>\n\u003Ch2>Getting Started\u003C\u002Fh2>\n\u003Cp>Install the latest stable release on macOS or Linux:\u003C\u002Fp>\n\u003Cpre>\u003Ccode class=\"language-bash\">curl -fsSL https:\u002F\u002Fapp.primeintellect.ai\u002Fprime-agent\u002Finstall.sh | sh\n\u003C\u002Fcode>\u003C\u002Fpre>\n\u003Cp>The installer downloads a versioned release, verifies its SHA-256 checksum, installs the \u003Ccode>prime-agent\u003C\u002Fcode> command, and can prepare the IPython runtime used by the agent.\u003C\u002Fp>\n\u003Cp>Start Prime Agent from the repository or directory you want it to work in:\u003C\u002Fp>\n\u003Cpre>\u003Ccode class=\"language-bash\">cd \u002Fpath\u002Fto\u002Fproject\nprime-agent\n\u003C\u002Fcode>\u003C\u002Fpre>\n\u003Cp>On first launch, run \u003Ccode>\u002Flogin\u003C\u002Fcode> to choose a subscription or API-key provider. Prime Agent works in the current directory a\u003C\u002Fp>\n",1786232078082]