Y Combinator Logo

Backed by Y Combinator

Build Reliable And Modular AI Agents

Drag-and-drop AI Agent Builder. Deploy to API in 1 click. Open-Source.

FROM LEADING AI INSTITUTIONS.

MetaBCGImperial CollegeYCJiosaavnUCLReka
MetaBCGImperial CollegeYCJiosaavnUCLReka
MetaBCGImperial CollegeYCJiosaavnUCLReka
MetaBCGImperial CollegeYCJiosaavnUCLReka

Features

All You Need For Reliable AI Agents

    Multimodal

    Works with Videos, Images, Audio, PDFs, etc.

    Modular Building Blocks

    Add self-refinement, knowledge, tools in seconds

    Profile Agent Behavior

    Observe what is going on in each step to intervene faster

    RAG

    Chunking, Parsing, Embedding, and Vector DB management

Why PySpur?

Built for Engineers

Deploy anywhere

Fully open-source with Apache2 license. Self-host on your own infrastructure. Keep in full control of your data.

bash

docker compose up --build -d

Easy to add your own tools

Adding new tools is a matter of simply creating a single Python file.

python

class MyNode(DynamicSchemaNode):
    name = "my_node"
    display_name = "MyNode"
    config_model = MyNodeConfig

    async def run(self, input_data: MyNodeInput) -> MyNodeOutput:
        ...

Easy sharing and version control

Every agent can be exported as a JSON file. Share with your team, and commit to git.

json

{
  "name": "Research Idea Generator",
  "definition": {
    "nodes": [
      {
        "id": "research information",
        "title": "research information",
        "node_type": "InputNode",
        "config": {
          "output_schema": {
            "existing_method_name": "str",
            "existing_method_code": "str",
            "existing_method_paper": "str",
            "additional_hints": "str",
            "goals": "str",
            "constraints": "str"
          },
          "enforce_schema": false,
          "title": "research information"
        },
        "coordinates": {
          "x": 0,
          "y": 0
        }
      }
    ]
  }
}

SDKs for TypeScript and Python

Integrate with your existing codebase and tools.

python

from pyspur import Agent

agent = Agent(
    id="S32",
)

response = agent.run(
    messages=[{"company": "Microsoft", "goal": "Buy or sell?"}]
)

Ready to get started?

Support us by starring our repository.