v1.7.321 shipped — see the roadmap

Multi-agent orchestration as a service.

Spin up swarms of specialized AI agents. Coordinate work across topologies. Multi-LLM out of the box. SaaS-ready, RLS-secured, built for teams.

5
LLM providers
15
Agents per swarm
4
Topology engines

Powering teams across

fintechhealthtechdevtoolsagenciesresearch labs

How it works

From signed in to first swarm in under five minutes.

  1. 01

    Connect your keys

    Bring your Claude / OpenAI / Gemini / Qwen / Ollama keys. Encrypted with AES-256-GCM at rest, decrypted only inside the request that needs them.

  2. 02

    Spin up a swarm

    Pick a topology, set a goal. Five specialized agents (architect, coder, tester, reviewer, security) split the work and report back through the chat UI.

  3. 03

    Ship continuously

    Track usage, swap models per workspace, upgrade tiers as you scale. Stripe-backed billing, customer portal, admin console — all built in.

Features

Built for production multi-agent workloads.

Swarm topologies

Hierarchical, mesh, adaptive, hierarchical-mesh. Match the architecture to the task.

Multi-LLM

Claude, GPT, Gemini, Qwen, Ollama. Switch per chat, per agent. Bring your own keys.

MCP tools

Memory, search, deployment, security. JSON-RPC over POST or stdio.

Workspace RLS

Postgres row-level security on every table. Multi-tenant by default.

GOAP planner

English goals → executable workflows. Visualize the plan tree.

Open dev

Track every release on /updates. Sourced from GitHub. Public.

What teams are saying

Angi gave us a real swarm in an afternoon. Topology presets are exactly the abstraction we kept reinventing.
MR
Maya R.
Staff Engineer · fintech
We swapped out three internal services for the MCP surface. Audit-friendly, RLS by default, BYO-keys means our compliance team didn't blink.
JL
Jonas L.
Eng Lead · healthtech
The /updates page sold our exec team. Watching the roadmap tick green in real time is more convincing than any sales deck.
PK
Priya K.
PM · enterprise SaaS

Pricing

Start free. Upgrade when you need more.

Free

Try the platform.

Free
  • 1 swarm · 3 agents
  • 50K tokens / month
  • Hierarchical topology
  • Community support
Start

Pro

Popular

For solo builders.

$29 /mo
  • 5 swarms · 10 agents/swarm
  • 1M tokens / month
  • Mesh + hierarchical
  • Email support
  • Custom GitHub OAuth
Start

Team

For growing teams.

$99 /mo
  • 20 swarms · 10 seats
  • 10M tokens / month
  • All four topologies
  • SSO + priority support
  • Audit log export
Start

Enterprise

For scale.

Custom
  • Unlimited everything
  • Custom models + on-prem
  • Dedicated support
  • SLA 99.9%
  • SOC 2 / DPA
Contact sales
PricingFreeProTeamEnterprise
Active swarms1520Unlimited
Agents per swarm3101550+
Monthly tokens50K1M10MUnlimited
TopologiesHierarchical+ MeshAll fourAll four
Workspace RLS
BYO API keys
Customer portal
SSO (SAML / Google)
Audit log export
Dedicated support
On-prem deployment

Frequently asked

Need something else? Talk to us.

  • Yes — Angi is a bring-your-own-keys platform. You configure each provider once at /settings/api-keys and the server decrypts them on a per-request basis. We never see plaintext.

  • Anthropic Claude (Opus / Sonnet / Haiku), OpenAI (GPT-5, GPT-4o), Google Gemini, Qwen 3.6 Max via DashScope, and Ollama for local models. You can switch model per chat or per agent.

  • Every table in Postgres has Row-Level Security enabled. A nightly CI job spins up two ephemeral tenants and asserts cross-workspace reads and writes are blocked.

  • Yes — Enterprise customers get a self-hosted or single-tenant deployment plus a DPA. The whole stack is open: Next.js + Supabase + Vue 3, no proprietary bits.

  • Stripe Checkout for monthly subscriptions, the customer portal for self-serve upgrades and invoices, and the workspace dashboard for live usage against your plan cap.

  • A way to wire agents together. Hierarchical = queen + leaves; mesh = peer; hierarchical-mesh = leaders mesh-peer with sub-clusters; adaptive = auto-pick by size. Each is one zod-schema-validated tool call away.

  • Yes. `angi login`, `angi chat "summarize this"`, `angi swarm init`, `angi usage`, `angi tools list` — all wired to the same MCP HTTP transport your chat UI uses. Tab-complete via `angi completion bash > /etc/bash_completion.d/angi`.

  • Yes. 1 swarm, 3 agents, 50K tokens/month. No credit card. Upgrade when you outgrow it.

Ready to ship multi-agent workflows?

Connect your LLM key, spin up your first swarm in under a minute.