Swarms API
The only API for orchestrating multi-agent systems at production scale.
Capabilities
Everything you need to ship agents in production
A complete multi-agent stack — orchestration, memory, tooling, and observability — behind one API.
Multi-agent orchestration
Sequential, concurrent, hierarchical, group-chat and custom topologies — handled by the runtime, not your code.
Shared memory
Persistent, scoped memory across agents so swarms can collaborate, branch, and reason over long horizons.
Pre-built tools
Web search, scraping, finance, document processing, code execution and dozens more — ready out of the box.
1,000+ models
OpenAI, Anthropic, Gemini, Groq, open-weight providers and hundreds more behind a single interface.
MCP protocol
Connect Notion, Supabase, GitHub or any internal system through the Model Context Protocol.
Streaming responses
Token-by-token streaming for low-latency UX and live progress on long-running swarm executions.
Rust runtime
Sub-millisecond initialization and zero-copy message passing built on a Rust-optimized core.
Enterprise infrastructure
Multi-region deployment, fault-tolerant retries, SOC 2 controls and encrypted transport by default.
Telemetry & logs
Per-agent traces, token accounting, latency metrics and structured logs for full observability.
Quickstart
Your first agent in three lines.
Sign up, drop in your API key, and call /v1/agent/completions. No infrastructure to provision.
Generate an API key on the platform
Install the SDK or call the REST endpoint
Run agents — we handle the runtime
import os
import requests
response = requests.post(
"https://api.swarms.world/v1/agent/completions",
headers={"x-api-key": os.environ["SWARMS_API_KEY"]},
json={
"agent_config": {
"agent_name": "Research Analyst",
"system_prompt": "You are a senior research analyst.",
"model_name": "gpt-4o",
"max_tokens": 8192,
"temperature": 0.5,
},
"task": "Summarize the state of multi-agent AI in 2025.",
},
)
print(response.json())Multi-agent
One request. A whole swarm.
Describe the topology and the agents — Swarms handles routing, memory, and aggregation.
import os
from swarms_client import SwarmsClient
client = SwarmsClient(api_key=os.environ["SWARMS_API_KEY"])
response = client.swarms.run(
name="Medical Analysis Swarm",
description="Analyzes symptoms and recommends ICD codes",
swarm_type="ConcurrentWorkflow",
task="45-year-old female with chest pain. Analyze and provide ICD codes.",
agents=[
{
"agent_name": "Symptom Analyzer",
"system_prompt": "You analyze patient symptoms and surface key indicators.",
"model_name": "gpt-4o",
"temperature": 0.3,
},
{
"agent_name": "ICD Code Specialist",
"system_prompt": "You map symptoms to ICD codes with rationale.",
"model_name": "gpt-4o",
"temperature": 0.2,
},
],
)
print(response)Available swarm topologies
Official SDKs
Native clients for every stack
Type-safe, idiomatic, production-ready libraries maintained by the Swarms team.
Pricing
Start free. Scale when you ship.
Generous free tier. Predictable team pricing. On-premise and enterprise when you need them.
Free
Get started with AI. No fees — perfect for prototyping.
Pay only for what you use
- 100 requests / minute
- 1,200 requests / day
- 200K tokens per agent
- 10 agents per batch
- Community support
Pro
For professionals who need more power and features.
per month
- Higher rate limits than Free
- Pay-as-you-go usage credits
- Priority queue access
- Email support
- All standard features
Premium
Production workloads with priority access.
per month · or $1,020/yr (save 15%)
- 2,000 requests / minute
- 100,000 requests / day
- 2M tokens per agent
- 500 agents per batch
- Priority availability zones
- 24/7 email & Discord support
Enterprise
Dedicated infrastructure, on-premise, and SLAs.
Tailored to your scale
- Custom rate limits
- Dedicated capacity
- On-premise deployment available
- Custom SLAs & integration support
- White-label options
FAQ
Common questions
Ready to ship a swarm?
Generate an API key in under a minute. Run your first multi-agent workflow before your coffee gets cold.