Developer hub
Build agents in 5 minutes.
Quick start, full documentation, API reference, and a thriving community. Everything you need to ship.
5-minute quick start
terminal
# 1. Install
npm install @cogstorm/core
# or: pip install cogstorm
# 2. Authenticate
cogstorm login
# 3. Scaffold
cogstorm init my-agent
# 4. Run locally
cd my-agent && cogstorm run flows/example.py
# 5. Ship
cogstorm deploy --productionHello world
hello.py
<span class="text-[#22D3EE]">from</span> cogstorm <span class="text-[#22D3EE]">import</span> <span class="text-[#A5B4FC]">Agent</span>, <span class="text-[#A5B4FC]">Orchestrator</span>
agent = <span <span class="text-[#22D3EE]">class</span>="text-[#67E8F9]"><span class="text-[#A5B4FC]">Agent</span></span><span <span class="text-[#22D3EE]">class</span>="text-[#94A3B8]">(</span>model=<span <span class="text-[#22D3EE]">class</span>="text-[#86EFAC]">"gpt-<span <span class="text-[#22D3EE]">class</span>="text-[#<span class="text-[#A5B4FC]">FCD34D</span>]">4</span>"</span>, tools=[<span <span class="text-[#22D3EE]">class</span>="text-[#86EFAC]">"web_search"</span>])
flow = <span <span class="text-[#22D3EE]">class</span>="text-[#67E8F9]"><span class="text-[#A5B4FC]">Orchestrator</span></span><span <span class="text-[#22D3EE]">class</span>="text-[#94A3B8]">(</span>agents=[agent])
result = <span class="text-[#22D3EE]">await</span> flow.<span <span class="text-[#22D3EE]">class</span>="text-[#67E8F9]">run</span><span <span class="text-[#22D3EE]">class</span>="text-[#94A3B8]">(</span><span <span class="text-[#22D3EE]">class</span>="text-[#86EFAC]">"<span class="text-[#A5B4FC]">What</span> are the top <span <span class="text-[#22D3EE]">class</span>="text-[#<span class="text-[#A5B4FC]">FCD34D</span>]">3</span> <span class="text-[#A5B4FC]">EV</span> trends?"</span>)API surface (preview)
A glimpse of the core primitives. Full reference in the docs.
| Symbol | Signature |
|---|---|
| Agent | Agent(name, model, tools, memory, instructions) |
| Tool | Tool.from_function(name, func) |
| Orchestrator | Orchestrator(agents, strategy, max_steps) |
| Flow.run | await flow.run(prompt) -> Result |
| Result | Result(output, traces, cost, tokens, duration) |
| Memory | Memory(backend, persistence) |