TypeScriptMITCreated Mar 31, 2026Last push 3d agoLatest release v1.24.0+11 stars this week+61 this month
Quick answers
What is open-multi-agent?
TypeScript orchestration framework that dynamically plans and executes tasks using multiple AI agents locally.
What does open-multi-agent do?
Open Multi-Agent shifts away from rigid, pre-defined execution graphs by utilizing a dynamic coordinator that plans task sequences at runtime based on high-level user goals. Written entirely in TypeScript, it acts as an orchestration layer that connects specialized AI agents to collaborate seamlessly. The framework is highly model-agnostic, supporting integration with external APIs like Claude and Gemini, as well as local execution via Ollama. This architecture allows developers to define what needs to be accomplished rather than explicitly hardcoding the entire execution path. It ensures privacy and full developer control by running complex multi-agent workflows locally.
Who is open-multi-agent for?
TypeScript developers and AI engineers building complex, goal-driven applications that require dynamic planning. It demands a solid understanding of language model capabilities and Node.js backend development.
How do I get started with open-multi-agent?
npm install @open-multi-agent/core
How popular is open-multi-agent on GitHub?
open-multi-agent/open-multi-agent has 7,069 stars and 2,435 forks on GitHub, and gained 11 stars in the last 7 days.
What license does open-multi-agent use?
open-multi-agent/open-multi-agent is released under the MIT license.
Star history
since Jul 28, 2026
7.1K stars as of Nov 19, 2026. Measured daily since Jul 28, 2026; GitHub no longer exposes earlier star timestamps.
Update history
5 recorded
Nov 15, 2026New release v1.24.0 (previously v1.23.0).
Nov 1, 2026New release v1.23.0 (previously v1.21.3).
Oct 20, 2026New release v1.21.3 (previously v1.21.1).
Oct 9, 2026New release v1.21.1 (previously v1.21.0).
Oct 4, 2026Previously tracked as JackChen-me/open-multi-agent; its 1 daily snapshot and 3 trending appearances were merged into this profile. Stars: 6,669 on 2026-07-28 under the old name, 6,972 on 2026-10-03 (+303).
Contribution activity
commits per day, last 52 weeks
573 commits in the last yearLessMore
Signals and awards
derived from tracked data
Very active
573 commits in 52 weeks
Well documented
High community health score
Permissive license
MIT
Continuous integration
Automated checks passing
Repeat trending
3 trending appearances
What open-multi-agent does
Open Multi-Agent shifts away from rigid, pre-defined execution graphs by utilizing a dynamic coordinator that plans task sequences at runtime based on high-level user goals. Written entirely in TypeScript, it acts as an orchestration layer that connects specialized AI agents to collaborate seamlessly. The framework is highly model-agnostic, supporting integration with external APIs like Claude and Gemini, as well as local execution via Ollama. This architecture allows developers to define what needs to be accomplished rather than explicitly hardcoding the entire execution path. It ensures privacy and full developer control by running complex multi-agent workflows locally.
TypeScript developers and AI engineers building complex, goal-driven applications that require dynamic planning. It demands a solid understanding of language model capabilities and Node.js backend development.
Dynamic execution planning: Automatically generates Directed Acyclic Graphs at runtime based on high-level goal descriptions.
Model-agnostic orchestration: Supports seamless integration with diverse language models including Claude, Gemini, and local Ollama.
Local privacy processing: Runs the entire multi-agent orchestration layer directly within the user's secure environment.
TypeScript architecture: Provides a robust, strongly-typed codebase for easy integration into existing Node.js backend services.
Specialized agent coordination: Manages state, communication, and tool access between disparate agents working on complex tasks.
Where teams use it
Dynamic workflow generation
Systems handle unpredictable user requests by mapping out custom execution steps on the fly without rigid programming.
Privacy-first processing
Enterprises run sensitive data analysis workflows entirely on local hardware using Ollama to ensure data protection.
Self-healing code pipelines
Developers deploy agents that can write code, run tests, and dynamically adjust their approach if errors occur.
Automated data synthesis
Users coordinate a swarm of specialized agents to independently gather and summarize information from multiple sources.
Agents your organization can own, approve, and audit.
OMA (Open Multi-Agent) is a self-hosted TypeScript agent runtime for coordinating agents, adding human approval gates, and inspecting runs offline.
No telemetry. No hosted control plane. Your keys, your models — cloud, local (Ollama, vLLM, llama-server), or Chinese providers — your environment. Nothing stops working when the people who built it leave.
Get started
Requires Node.js 20 or newer. For production, use a currently maintained
Node.js LTS release. Node.js 20 is upstream-EOL and retained only as a
migration compatibility window; OMA will remove it in the next major release,
no earlier than 2026-10-31.
Scaffold a PR review agent, security analysis agent, or teaching DAG:
npm create oma-app@latest my-oma
In an interactive terminal, that one command selects a starter and runtime, installs dependencies, and runs a deterministic local demo. The demo needs no API key and makes no model request: scripted model responses drive the real OMA scheduler, result aggregation, and offline dashboard.
Or add OMA to an existing backend:
npm install @open-multi-agent/core
import{OpenMultiAgent}from'@open-multi-agent/core'constoma=newOpenMultiAgent({defaultProvider: 'openai',defaultModel: 'gpt-5.4'})constteam=oma.createTeam('research-team',{name: 'research-team',agents: [{name: 'researcher',systemPrompt: 'Find the relevant facts.'},{name: 'analyst',systemPrompt: 'Compare evidence and identify tradeoffs.'},],sharedMemory: true,})constresult=awaitoma.runTeam(team,'Compare three approaches and recommend one.')// Nothing above declares a task graph. The coordinator planned one at runtime,// and the finished run is data you can read back.for(consttaskofresult.tasks??[]){console.log(`[${task.status}] ${task.title} → ${task.assignee??'unassigned'}`,task.dependsOn)}console.log(result.agentResults.get('coordinator')?.output)console.log(result.totalTokenUsage)
Set OPENAI_API_KEY to run this example. Providers covers other hosted models, local servers, OpenAI-compatible endpoints, and AI SDK providers.
runAgent() runs a single agent, runTasks() executes an explicit pipeline, and runTeam() plans from a goal. The Core package guide walks through all three modes, provider and credential setup, and the production checklist. The example index lists every runnable example across basics, cookbook workflows, patterns, providers, and integrations.
Durable approvals
Pause before executing a plan, dispatching a task, or calling a tool for human approval. With persistent storage, a restarted process can continue from the reviewed content. Your application supplies the approval interface and notifications. Guide
Verifiable journal
Enable a run journal to record model inputs, tool calls, and results for later inspection. verifyRun() checks the recorded sources and content for consistency offline; it does not prove the journal was never edited. Guide
Governance floor
Declare required roles and their execution order, then check whether the actual run met those requirements. A successful run can still leave governance requirements unmet, so applications must check governanceConclusion separately. Guide
Runs where you run
No telemetry, no hosted control plane. A library with no OMA backend or account, and none planned. It makes no analytics, license, update, or phone-home request. Self-hosting
Your keys, your models. Built-in adapters for Anthropic, OpenAI, Azure OpenAI, Bedrock, Gemini, Grok, NVIDIA, and Copilot, and for DeepSeek, Doubao, Hunyuan, MiniMax, MiMo, and Qiniu; Ollama, vLLM, and llama-server through baseURL; any OpenAI-compatible endpoint and Vercel AI SDK providers. Providers
Egress policy.offline or allowlist, checked before a built-in adapter connects. A child policy can only tighten its parent, an unenforceable transport fails closed, and process and ACP backends sit outside it. LLM egress policy
Built with OMA
open-multi-agent launched 2026-04-01 under MIT. Known users and integrations to date:
temodar-agent by Ali Sünbül. WordPress security analysis platform running OMA's built-in tools (bash, file_*, grep) inside a Docker runtime. Confirmed production use.
Mark Galyan runs OMA fully offline on local quantized models, using the coordinator and context compaction to keep autonomous agent loops alive under tight VRAM limits. Contributor since the framework's first month.
Engram: "Git for AI memory." Syncs knowledge across agents instantly and flags conflicts. (repo)
More users and integrations
Users
PR-Copilot by kidoom. AI pull-request review assistant running an OMA review team, with defineTool repo-context tools and a custom ContextStrategy for token-aware diff compression.
StuFlow by znc15. Terminal AI coding assistant on OMA's orchestration core, driving runAgent / runTasks / runTeam with a custom coordinator, paired with DeepSeek.
Reports to Charts Studio. Turns documents and research tables into slide-ready charts, using a five-role extraction council with structured outputs and deterministic validation.
OrchestrAI: Visual multi-agent workspace using OMA for task-graph execution and peer review, with PostgreSQL-backed checkpoints for restart recovery.
Integrations
@agentsonar/oma: Sidecar detecting cross-run delegation cycles, repetition, and rate bursts.
CodingScaffold: Agentic-coding scaffold that lists OMA as an optional orchestration backend, with a runTeam workflow template.
baize-oma: HTTP adapter exposing OMA runAgent() and runTeam() as Baize slot capabilities.
We build customer-owned systems on OMA for organizations that need one. Email [email protected].
Sponsors
Paid sponsors supporting open-multi-agent. Sponsorship does not affect technical decisions or model recommendations.
Providers
Atlas Cloud: Full-modal AI inference platform giving one API for video, image, and LLM across 300+ curated models. $5 credit vouchers for OMA users, first come first served. See the Atlas Cloud setup guide.
Optional coordinator
runTeam() decomposes a goal into a task graph across agents. One model call turns the goal into task specs with assignees and dependencies, a deterministic scheduler executes them, and a second call writes the final answer from the completed task outputs. The coordinator is never consulted mid-run, and the finished run is data you can read back. Use runAgent() or runTasks() when you already know the work.
The offline Run Viewer replaying a real run from the trace store: task DAG, span waterfall, and per-task evidence, with no hosted service involved.
Coordinator covers what it decides and what it is allowed to see. Plan replay freezes an approved plan, Consensus verifies outputs with independent judges, and External agents puts Claude Code, Gemini CLI, and Codex on the same task graph through process and ACP backends.
Packages
@open-multi-agent/core: Runtime, tools, memory, checkpoints, approvals, journal, traces, CLI, and offline Run Viewer.
@open-multi-agent/otel: Optional OpenTelemetry adapter for teams with a centralized OpenTelemetry stack.
create-oma-app: Scaffolder behind npm create oma-app; starter templates with a no-key local demo.
Core users can store traces locally and inspect them with the offline Run Viewer. Install the OTel package only when OMA traces should appear in the same monitoring system as the rest of your application.
### Added
- Added an experimental createImageTool so a model can call runImage to generate or edit images from a tool call.
- Added the nvidia provider for NVIDIA NIM models, defaulting to the hosted NIM endpoint and falling back to the NVIDIA API key environment variable, with base URL support for self-hosted deployments.
### Changed
- Claude Opus 4.7, Sonnet 5, and later models now omit the sampling parameters temperature, top_p, and top_k, which those models reject, on the Anthropic adapter and on Bedrock (foundation-model ids, inference profiles, and ARNs). Older Claude models and non-Claude Anthropic-compatible models keep their sampling values.
- The provider template now suggests current models instead of retired ones: an Azure deployment-name placeholder, a current Haiku for Bedrock, and a current Gemini flash model.
- Provider safety and guardrail finish reasons now map to content_filter: Bedrock guardrail-intervened and content-filtered, and Gemini safety, recitation, blocklist, prohibited-content, SPII, and image-safety finishes.
### Fixed
- Tool results remain contiguous when sliding-window and summarize strategies rewrite history, and split results stay with their
### Breaking changes
This release ships as a minor version, so a caller on a `^1.x` range receives it without an explicit upgrade step. A project that pins zod below 3.25, code that uses object-only members of `taskProfileSchema` or `gatePolicySchema`, and a coordinator plan over the new size caps can stop working after the upgrade. The individual entries are under Changed below.
- **`zod` is now a required peer dependency, `^3.25.0 || ^4.6.6`.** A project that depends on zod 3.23 or 3.24 has to upgrade zod first; until it does, npm stops with a peer dependency conflict. With a package manager that does not install peers automatically, add `zod` to your own dependencies.
- **`taskProfileSchema` and `gatePolicySchema` are typed as `ZodSchema<T>`.** Object-only members such as `.extend()` and `.shape` no longer type-check on them. Define your own schema where you extended them.
- **Coordinator plans are capped.** A plan with more than 200 tasks, a task with more than 200 `dependsOn` entries, or more than 64 `requires.requiredTools` or `requires.requiredCapabilities` entries now fails validation.
### Added
- Zod 4 support. Tool `inputSchema` and `outputSchema`, `AgentConfig.output
### Breaking changes
This release ships as a minor version, so a caller on a `^1.x` range receives it without an explicit upgrade step. A caller that passes malformed task metadata or disables thinking on MiniMax-M3.1-Flash still installs, then fails when the code runs. The individual entries are under Changed and Fixed below.
- **`Team.addTask()` now honors every task option it accepts.** Options that were previously discarded — retry settings, memory scope, dependency payload, role, priority, metadata, capability requirements, and consensus verification — now affect queue ordering, retries, capability gating, memory scope, and consensus verification. Malformed metadata that was previously ignored now fails validation before the task is enqueued.
- **MiniMax-M3.1-Flash cannot disable thinking.** A call with `thinking: { enabled: false }` now throws before the request, where earlier versions forwarded it with thinking ignored. Use MiniMax-M3 when disabled thinking is required, or pass an explicit `extraBody.thinking` override.
### Added
- MiniMax-M3.1-Flash support, including adaptive thinking, reasoning-effort forwarding, and native reasoning replay on tool-call follow-ups.
-
### Fixed
- Symlink loops and chains longer than the kernel's link limit now fail the sandbox check with a `Too many levels of symbolic links` error. Earlier versions also failed these writes, with the filesystem error instead.
- A symlink target that ends in a separator or `.` must now name an existing directory, as the kernel requires. On macOS, earlier versions resolved such a link to the file of that name, so `file_write` overwrote it.
### Security
- `file_write` can no longer create a file outside the agent's working directory through a dangling symlink (GHSA-h42w-g3ch-rjjm, CVE-2026-108600). Versions 1.5.0 through 1.21.2 are affected. The sandbox check now resolves a dangling symlink to the target it names, one path component at a time as the kernel does, so a link pointing outside the root is rejected like any other outside path and a link pointing inside it keeps working. Every built-in filesystem tool shares this check.
### Compatibility
- No public export, signature, or error `ToolResult` shape changed; upgrading needs no code change.
- Symlinks that resolve inside the working directory work as before, including a dangling link whose in-root target `file_write` creat
### Changed
- Updated the core package README, which ships in the core tarball, to route community feedback to GitHub Issues instead of Discussions.
### Compatibility
- No public API, runtime, provider, template, or published dependency change is present; upgrading requires no caller action.
- create-oma-app templates will pin the new core patch exactly, keeping scaffolded projects on a consistent core version.
- Internal release tooling only: release-bot defaults now allow up to 48 registry polls at 10s per package instead of 9. Maintainers needing the old fail-fast behavior can still pass pollAttempts and pollDelayMs explicitly; this package is private and never published.
- If the reported partial publication of the previous release is confirmed, consumer action is still not required; recovery is a maintainer dispatch of publish.yml on the exact merged release commit reachable from main with a successful CI run.
## Packages
- `@open-multi-agent/core`: `1.21.2`
- `@open-multi-agent/otel`: remains at `0.1.4` and is not republished
- `create-oma-app`: `0.8.9`; generated starters pin core `1.21.2`
## Install
```bash
npm i @open-multi-agent/[email protected]
npm create oma-app@lates
Code frequency
additions and deletions
+280K lines added, -114.5K removed over the last year.
Commits per week
last 52 weeks
573 commits in the last 52 weeks.
When work happens
weekday and hour
Commit volume by weekday and hour (UTC). Larger dots mean more commits.
The agent harness performance optimization system. Skills, instincts, memory, security, and research-first development for Claude Code, Codex, Opencode, Cursor and beyond.