OpenSwarm

Autonomous AI agent orchestrator — Claude, GPT, Codex, OpenRouter (any model), and local models (Ollama/LMStudio/llama.cpp)

💬 Help shape OpenSwarm. Share feature ideas, vote on the roadmap, and ask questions in GitHub Discussions. The roadmap is built in the open — your feedback decides what ships next.

OpenSwarm orchestrates multiple AI agents as autonomous code workers. It picks up Linear issues, runs Worker/Reviewer pair pipelines, reports to Discord, and retains long-term memory via LanceDB. Workers run on Claude Code, OpenAI GPT, Codex, any OpenRouter model, or local open-source models (Ollama, LMStudio, llama.cpp) — with cost-aware routing measured on an L0–L6 benchmark ladder.

Verified on real GitHub issues: the agentic harness solves SWE-bench Lite instances graded by the official harness. Hybrid mode — a frontier model diagnoses read-only, a lightweight model implements with a verification loop — resolved 3/3 attempted instances that every single lightweight model had failed, at a fraction of frontier-only cost. Workers also learn each repository over time: task outcomes are stored as per-repo knowledge and recalled into future prompts. (benchmark rubric & results)

Quick Start

bash
npm install -g @intrect/openswarm
openswarm

That's it. openswarm with no arguments launches the TUI chat interface immediately.

TUI Chat Interface

TUI keyboard shortcuts

Key	Action
`Tab`	Switch tabs (Chat / Projects / Tasks / Stuck / Logs)
`Enter`	Send message
`Shift+Enter`	Newline
`i`	Focus input
`Esc`	Exit input focus
`Ctrl+C`	Quit

Status bar shows: provider · model · message count · cumulative cost

CLI Commands

bash
openswarm                        # TUI chat (default)
openswarm chat [session]         # Simple readline chat
openswarm start                  # Start full daemon (requires config.yaml)
openswarm run "Fix the bug" -p ~/my-project   # Run a single task
openswarm exec "Run tests" --local --pipeline # Execute via daemon
openswarm init                   # Generate config.yaml scaffold
openswarm validate               # Validate config.yaml

# Code Registry & BS Detector
openswarm check --scan           # Scan repo → register all entities
openswarm check src/foo.ts       # File brief (entities, tests, risk)
openswarm check --bs             # BS pattern scan (bad code smells)
openswarm check --stats          # Registry statistics
openswarm check --high-risk      # High-risk entities
openswarm check --search "name"  # Full-text search
openswarm annotate "funcName" --deprecate "reason"
openswarm annotate "funcName" --tag "needs-refactor"
openswarm annotate "funcName" --warn "error/security: SQL injection"

`openswarm exec` options

Option	Description
`--path <path>`	Project path (default: cwd)
`--timeout <seconds>`	Timeout in seconds (default: 600)
`--local`	Execute locally without daemon
`--pipeline`	Full pipeline: worker + reviewer + tester + documenter
`--worker-only`	Worker only, no review
`-m, --model <model>`	Model override for worker

Exit codes: 0 success · 1 failure · 2 timeout

Full Daemon Setup

For autonomous operation (Linear issue processing, Discord control, PR auto-improvement), you need a full config:

Prerequisites

Node.js >= 22
Claude Code CLI authenticated (claude -p) — default provider
OpenAI Codex CLI (codex exec) — optional alternative provider
Discord Bot token with message content intent
Linear API key and team ID
GitHub CLI (gh) for CI monitoring (optional)

Configuration

bash
git clone https://github.com/unohee/OpenSwarm.git
cd OpenSwarm
npm install
cp config.example.yaml config.yaml

Create a .env file:

bash
DISCORD_TOKEN=your-discord-bot-token
DISCORD_CHANNEL_ID=your-channel-id
LINEAR_API_KEY=your-linear-api-key
LINEAR_TEAM_ID=your-linear-team-id

config.yaml supports ${VAR} / ${VAR:-default} substitution and is validated with Zod schemas.

Key configuration sections

Section	Description
`discord`	Bot token, channel ID, webhook URL
`linear`	API key, team ID
`github`	Repos list for CI monitoring
`agents`	Agent definitions (name, projectPath, heartbeat interval)
`autonomous`	Schedule, pair mode, role models, decomposition settings
`prProcessor`	PR auto-improvement schedule, retry limits, conflict resolver config

CLI Adapter (Provider)

yaml
adapter: claude   # "claude" | "codex" | "gpt" | "openrouter" | "local" | "lmstudio"

Switch at runtime via Discord: !provider codex / !provider claude

Adapter	Backend	Models	Auth
`claude`	Claude Code CLI	sonnet-4, haiku-4.5, opus-4	CLI auth
`codex`	OpenAI Codex CLI	o3, o4-mini	CLI auth
`gpt`	OpenAI API	gpt-4o, o3, gpt-4.1	OAuth PKCE
`openrouter`	OpenRouter API (native agentic loop)	any OpenRouter model — gpt-5, gemini-2.5-flash, deepseek, glm, qwen, …	OAuth PKCE or `OPENROUTER_API`
`local`	Ollama / LMStudio / llama.cpp	gemma4, llama3, mistral, qwen, etc.	None
`lmstudio`	LM Studio OpenAI-compatible API	loaded LM Studio model (`LMSTUDIO_MODEL`)	Optional API key

The openrouter adapter runs OpenSwarm's own agentic tool loop (read/search/edit/bash with verification guards), enables ZDR (data_collection: deny) for non-OpenAI models, and applies Anthropic prompt caching automatically.

Local models are auto-detected on standard ports (Ollama :11434, LMStudio :1234, llama.cpp :8080). Use lmstudio for a dedicated LM Studio endpoint (LMSTUDIO_BASE_URL, default http://localhost:1234).

Per-role adapter overrides:

yaml
autonomous:
  defaultRoles:
    worker:
      adapter: codex
      model: o4-mini
    reviewer:
      adapter: claude
      model: claude-sonnet-4-20250514

Agent Roles

yaml
autonomous:
  defaultRoles:
    worker:
      model: claude-haiku-4-5-20251001
      escalateModel: claude-sonnet-4-20250514
      escalateAfterIteration: 3
      timeoutMs: 1800000
    reviewer:
      model: claude-haiku-4-5-20251001
      timeoutMs: 600000
    tester:
      enabled: false
    documenter:
      enabled: false
    auditor:
      enabled: false

Running the daemon

macOS launchd service (recommended)

bash
npm run service:install    # Build and install as system service
npm run service:start      # Start
npm run service:stop       # Stop
npm run service:restart    # Restart
npm run service:status     # Status and recent logs
npm run service:logs       # stdout (follow mode)
npm run service:errors     # stderr (follow mode)
npm run service:uninstall  # Uninstall

Manual

bash
npm run build && npm start   # Production
npm run dev                  # Development (tsx watch)
docker compose up -d         # Docker

Architecture

                         ┌──────────────────────────┐
                         │       Linear API          │
                         │   (issues, state, memory) │
                         └─────────────┬────────────┘
                                       │
                 ┌─────────────────────┼─────────────────────┐
                 │                     │                     │
                 v                     v                     v
  ┌──────────────────┐  ┌──────────────────┐  ┌──────────────────┐
  │ AutonomousRunner │  │  DecisionEngine  │  │  TaskScheduler   │
  │ (heartbeat loop) │─>│  (scope guard)   │─>│  (queue + slots) │
  └────────┬─────────┘  └──────────────────┘  └────────┬─────────┘
           │                                            │
           v                                            v
  ┌──────────────────────────────────────────────────────────────┐
  │                      PairPipeline                            │
  │  ┌────────┐   ┌──────────┐   ┌────────┐   ┌─────────────┐  │
  │  │ Worker │──>│ Reviewer │──>│ Tester │──>│ Documenter  │  │
  │  │(Adapter│<──│(Adapter) │   │(Adapter│   │  (Adapter)  │  │
  │  └───┬────┘   └──────────┘   └────────┘   └─────────────┘  │
  │      │  ↕ StuckDetector                                      │
  │  ┌───┴────────────────────────────────────────────────────┐  │
  │  │ Adapters: Claude | Codex | GPT | Local (Ollama/LMS)   │  │
  │  └────────────────────────────────────────────────────────┘  │
  └──────────────────────────────────────────────────────────────┘
           │                     │                     │
           v                     v                     v
  ┌──────────────┐  ┌──────────────────┐  ┌──────────────────┐
  │  Discord Bot │  │  Memory (LanceDB │  │  Knowledge Graph │
  │  (commands)  │  │  + Xenova E5)    │  │  (code analysis) │
  └──────────────┘  └──────────────────┘  └────────┬─────────┘
                                                    │
                                           ┌────────┴─────────┐
                                           │  Code Registry   │
                                           │  (SQLite + FTS5) │
                                           │  + BS Detector   │
                                           └──────────────────┘

Features

Multi-Provider Adapters — Pluggable adapter system: Claude Code, OpenAI GPT/Codex, OpenRouter (any model, native agentic loop), and local models (Ollama, LMStudio, llama.cpp) with runtime provider switching
Code Registry — SQLite-backed entity registry tracking every function/class/type across 8 languages, with complexity scoring, test mapping, and risk assessment
BS Detector — Built-in static analysis engine that detects bad code patterns (empty catch, hardcoded secrets, as any, etc.) with pipeline guard integration
Autonomous Pipeline — Cron-driven heartbeat fetches Linear issues, runs Worker/Reviewer pair loops, and updates issue state automatically
Worker/Reviewer Pairs — Multi-iteration code generation with automated review, testing, and documentation stages
Decision Engine — Scope validation, rate limiting, priority-based task selection, and workflow mapping
Cognitive Memory — LanceDB vector store with Xenova/multilingual-e5-base embeddings for long-term recall across sessions
Repo Knowledge Loop — workers learn each repository over time: task outcomes (success patterns, review-rejection pitfalls) are stored per-repo and recalled into the next worker prompt
SWE-bench Verified — the agentic harness solves real SWE-bench Lite issues, graded by the official harness; hybrid mode (frontier diagnosis + lightweight implementer) resolved 3/3 attempted instances (benchmarks/RUBRIC.md)
Knowledge Graph — Static code analysis, dependency mapping, impact analysis, and file-level conflict detection across concurrent tasks
Discord Control — Full command interface for monitoring, task dispatch, scheduling, provider switching, and pair session management
Rich TUI Chat — Claude Code inspired terminal interface with tabs, streaming responses, and geek-themed loading messages
Dynamic Scheduling — Cron-based job scheduler with Discord management commands
PR Auto-Improvement — Monitors open PRs, auto-fixes CI failures, auto-resolves merge conflicts, and retries until all checks pass
Long-Running Monitors — Track external processes (training jobs, batch tasks) and report completion
Web Dashboard — Real-time pipeline stages, cost tracking, worktree status, and live logs on port 3847
Pace Control — 5-hour rolling window task caps, per-project limits, turbo mode, exponential backoff on failures
i18n — English and Korean locale support

How It Works

Linear (Todo/In Progress)
  → Fetch assigned issues
  → DecisionEngine filters & prioritizes
  → Resolve project path via projectMapper
  → PairPipeline.run()
    → Worker generates code (Claude CLI)
    → Reviewer evaluates (APPROVE/REVISE/REJECT)
    → Loop up to N iterations
    → Optional: Tester → Documenter stages
  → Update Linear issue state (Done/Blocked)
  → Report to Discord
  → Save to cognitive memory

Memory System

Hybrid retrieval: 0.55 × similarity + 0.20 × importance + 0.15 × recency + 0.10 × frequency

Memory types: belief · strategy · user_model · system_pattern · constraint

Background: decay, consolidation, contradiction detection, distillation.

Repo knowledge loop — every completed task writes repo-scoped knowledge
(success → system_pattern with files changed + approach, review rejection →
constraint pitfall), and the next task on the same repo recalls the most
relevant entries into the worker prompt as a "Repository Knowledge" section.
Workers get better at a codebase the more they work on it.

Benchmarks (L0–L6)

benchmarks/ contains a difficulty ladder for routing models by measured
capability — synthetic L0–L5 tasks with deterministic grading, and L6 = real
GitHub issues (SWE-bench Lite) solved by the OpenSwarm harness and graded by
the official swebench harness. Headline: hybrid mode (frontier read-only
diagnosis + lightweight implementer with a verification loop) resolved 3/3
attempted instances that every single lightweight model had failed. See
benchmarks/RUBRIC.md for the rubric, measured results,
and the harness defects the benchmark uncovered.

Discord Commands

Task Dispatch

Command	Description
`!dev <repo> "<task>"`	Run a dev task on a repository
`!dev list`	List known repositories
`!tasks`	List running tasks
`!cancel <taskId>`	Cancel a running task

Agent Management

Command	Description
`!status`	Agent and system status
`!pause <session>`	Pause autonomous work
`!resume <session>`	Resume autonomous work
`!log <session> [lines]`	View recent output

Linear Integration

Command	Description
`!issues`	List Linear issues
`!issue <id>`	View issue details
`!limits`	Agent daily execution limits

Autonomous Execution

Command	Description
`!auto`	Execution status
`!auto start [cron] [--pair]`	Start autonomous mode
`!auto stop`	Stop autonomous mode
`!auto run`	Trigger immediate heartbeat
`!approve` / `!reject`	Approve or reject pending task

Worker/Reviewer Pair

Command	Description
`!pair`	Pair session status
`!pair start [taskId]`	Start a pair session
`!pair run <taskId> [project]`	Direct pair run
`!pair stop [sessionId]`	Stop a pair session
`!pair history [n]`	View session history
`!pair stats`	View pair statistics

Scheduling

Command	Description
`!schedule`	List all schedules
`!schedule run <name>`	Run a schedule immediately
`!schedule toggle <name>`	Enable/disable a schedule
`!schedule add <name> <path> <interval> "<prompt>"`	Add a schedule
`!schedule remove <name>`	Remove a schedule

Other

Command	Description
`!ci`	GitHub CI failure status
`!provider <claude\|codex>`	Switch CLI provider at runtime
`!codex`	Recent session records
`!memory search "<query>"`	Search cognitive memory
`!help`	Full command reference

Project Structure

src/
├── index.ts                 # Entry point
├── cli.ts                   # CLI entry point (run, exec, chat, init, validate, start)
├── cli/                     # CLI subcommand handlers
│   └── promptHandler.ts     # exec command: daemon submit, auto-start, polling
├── core/                    # Config, service lifecycle, types, event hub
├── adapters/                # CLI provider adapters (claude, codex, gpt, local), process registry
├── agents/                  # Worker, reviewer, tester, documenter, auditor
│   ├── pairPipeline.ts      # Worker → Reviewer → Tester → Documenter pipeline
│   ├── agentBus.ts          # Inter-agent message bus
│   └── cliStreamParser.ts   # Claude CLI output parser
├── orchestration/           # Decision engine, task parser, scheduler, workflow
├── automation/              # Autonomous runner, cron scheduler, PR processor
├── memory/                  # LanceDB + Xenova embeddings cognitive memory
├── knowledge/               # Code knowledge graph (scanner, analyzer, graph)
├── registry/                # Code entity registry, BS detector, entity scanner
├── issues/                  # Local issue tracker (SQLite + GraphQL + Kanban UI)
├── discord/                 # Bot core, command handlers, pair session UI
├── linear/                  # Linear SDK wrapper, project updater
├── github/                  # GitHub CLI wrapper for CI monitoring
├── support/                 # Web dashboard, planner, rollback, git tools
├── locale/                  # i18n (en/ko) with prompt templates
└── __tests__/               # Vitest test suite

State & Data

Path	Description
`~/.openswarm/`	State directory (memory, codex, metrics, workflows)
`~/.openswarm/registry.db`	Code entity registry (SQLite)
`~/.openswarm/issues.db`	Local issue tracker (SQLite)
`~/.claude/openswarm-*.json`	Pipeline history and task state
`config.yaml`	Main configuration
`dist/`	Compiled output

Tech Stack

Category	Technology
Runtime	Node.js 22+ (ESM)
Language	TypeScript (strict mode)
Build	tsc
Agent Execution	Claude Code, OpenAI GPT/Codex, Ollama/LMStudio/llama.cpp
Local DB	better-sqlite3 (WAL mode, FTS5)
Task Management	Linear SDK (`@linear/sdk`)
Communication	Discord.js 14
Vector DB	LanceDB + Apache Arrow
Embeddings	Xenova/transformers (multilingual-e5-base, 768D)
Scheduling	Croner
Config	YAML + Zod validation
Linting	oxlint
Testing	Vitest

Changelog

v0.3.0

Code Registry: openswarm check --scan scans repo, registers 1000+ entities across 8 languages (TS, Python, Go, Rust, Java, C, C++, C#) with test mapping, complexity scoring, and risk assessment
BS Detector: openswarm check --bs — built-in static analysis for bad code patterns, pipeline guard integration
Local Model Support: Ollama, LMStudio, llama.cpp via single local adapter with auto-detection
GPT Adapter: OpenAI models via OAuth PKCE flow
Local Issue Tracker: SQLite + GraphQL + Kanban web UI at :3847/issues
CLI: openswarm check, openswarm annotate commands

v0.2.2

openswarm without arguments now launches TUI chat directly

v0.2.1

Security: patched lodash, picomatch, rollup, undici, yaml vulnerabilities

v0.2.0

Published as @intrect/openswarm on npm
Extracted @intrect/claude-driver as standalone zero-dependency package
Autonomous runner hardening and multi-project orchestration
Task-state rehydration from Linear comments
--verbose flag for detailed execution logging
Codex adapter: dropped o-series model override

v0.1.0

Initial release
Worker/Reviewer pair pipeline
Claude Code CLI + Codex CLI adapters
Discord bot control
Linear integration
LanceDB cognitive memory
Web dashboard (port 3847)
Rich TUI chat interface

License

GPL-3.0