PrimeIntellect-ai /prime-agent
A self-improving RLM agent for coding workflows and long-running autonomous tasks.
★ 22,709⑂ 2,516👁 85⑃ 3,041● 459▤ 61
+6
Daily
+156
Weekly
+754
Monthly
22,709
stars
2,516
forks
85
watchers
3,041
prs
459
issues
61
releases
README
Prime Agent: A Self-Improving RLM Harness
Documentation •
Verifiers •
PRIME-RL
Prime Agent is an open-source coding and research agent for general and long-running work. It is designed around two core abstractions:
- The Recursive Language Model (RLM) treats context as variables (prompt-as-a-variable) and tools like recursive subagents as function calls (programmatic tool /sub-agent calling) inside a persistent REPL.
- The Continual Harness 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.
Install
Install the latest stable version with the one-command installer:
bash
curl -fsSL https://app.primeintellect.ai/prime-agent/install.sh | sh
On Windows, install from PowerShell:
powershell
irm https://app.primeintellect.ai/prime-agent/install.ps1 | iex
The Windows installer downloads the latest stable Windows x64 build and verifies its checksum before installation.
Why Prime Agent
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.
- Everything is programmatic: a persistent Python REPL is the built-in model tool; file operations, shell commands, tool use, subagents, and context management happen through code.
- Subagents are built in: rlm.spawn(...) spawns real child agents for parallel or background work and returns their results programmatically.
- The harness can improve: /refine 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.
- Skills can be executable: Python skills in your user configuration are imported into the kernel; repository skills provide Markdown instructions without automatic code execution.
- 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.
Slack
Connect Slack from /mcp or run /mcp login slack inside Prime Agent to search
your workspace and use Slack tools. See the Slack connector guide
for OAuth setup and troubleshooting.
Google Drive and Google Sheets
Connect Google's official Drive and Sheets MCP servers with your own Google
OAuth client: add them to mcpServers with clientId, clientSecret,
callbackPort, and scopes, then run /mcp login . See the
Google Drive and Google Sheets guide for the
Google Cloud setup and the tool mappingTop issues
#3315Not possible to paste api key in login process3 comments · 2026-10-02#3250TUI drops the second of two identical characters that arrive together (key repeat over ssh, dictation, IME)3 comments · 2026-10-01#3174pa-tui: update_status_row panics on an out-of-range index (prepare_entry_mutation runs before the status-row guard)3 comments · 2026-09-30#1811fix: handle empty env vars in config (prevent relative path bug)3 comments · 2026-08-27#33140.9.6: resume refused after bridge termination; failed descriptor retains ownerClientId2 comments · 2026-10-02#3176pa-tui: key_event_to_id collapses modifier-bearing special keys (ctrl+tab -> tab, ctrl+space -> 'ctrl+ ') so their bindings never match2 comments · 2026-09-30#3739watch-agent: shared per-child feed needs resolved-id keying and cross-consumer refcounting1 comments · 2026-10-22#3312pa-daemon failover: routed image episodes re-send 'backup' retries to the failed provider (switch keeps route.target), plus mid-turn set_model race1 comments · 2026-10-02
Discussions
#4067[Feature] Keenable as a web search provider that works without a key0 comments · 2026-11-14#4059[Feature] Terminal program status reporting over OSC 75010 comments · 2026-11-13#4054[Feature] TUI: support custom themes (or per-color overrides); warn when a theme name is unknown0 comments · 2026-11-12#4049Support OSC 7501 agent status reporting0 comments · 2026-11-12#4028Feature Request: Portable Agent Workspaces — Export and Import Complete Sessions Across Machines0 comments · 2026-11-11#4027[v0.10.3] Full ACP compatibility for bb, Zed, JetBrains, and all ACP-compatible editors1 comments · 2026-11-11#4020[Bug] rlm.collect(timeout_ms>0) returns running children at once in daemon sessions (held task prompt)0 comments · 2026-11-10#4018[Bug] Claude Fable 5.2 blocked on subscription auth: Claude Code version gate moved to 2.1.2903 comments · 2026-11-09