An API for AI agent workflows
Seamlessly connect any AI model, framework, or agent. Sync data, coordinate actions, and trigger events in real-time—with one universal, lightweight REST API.
# Chain AI workflows in just a few lines
import gumm
gumm.send(
context="my-project",
sender="agent:researcher",
recipients=["gpt-4"],
data={"action": "Summarize", "report": "..."}
)
@gumm.on_webhook(event="message.created")
def handle_response(message):
# Trigger agent:analyst to action
Build with any framework.
Connect with one API.
Connect Any AI Model or Agent
Bridge OpenAI, Claude, LangChain, AutoGen, or custom frameworks. No refactoring—wrap existing agents in minutes with Python/JS/PHP SDKs.
Event-Driven Workflows
Trigger agents via webhooks when new messages arrive. Chain agents dynamically (Agent A → Agent B) or pause for human input.
Context-Aware Collaboration
Group messages by context for conversations, projects, or workflows. Guide agents with metadata (e.g., "intent": "summarize")—no rigid schemas.
Decentralized & Scalable
Agents communicate directly or via pub-sub—no central orchestrator. Stateless API scales automatically on serverless, containers, or bare metal.
Stick it together with Gumm
Build dynamic workflows, from simple chains to complex human-in-the-loop systems—all powered by a single API.
Multi-Agent Workflows
Example: Research agent → Summarizer → Report writer. How: Pass data between agents using context_id and JSON.
Human-in-the-Loop Automation
Approve/reject AI outputs before agents proceed. Audit trails—Every message is timestamped & queryable via API.
Rapid Experimentation
Swap models (e.g., GPT-3.5 → GPT-4) by changing one endpoint. No pipeline rewrites—test new agents in minutes.
Connect Old & New
Bridge Python AI tools, JavaScript chatbots, and PHP CMS systems. JSON over REST works everywhere.
Why choose Gumm
Works with Your Stack
Use REST, webhooks, and SDKs easily within your existing code. No new infrastructure required.
Zero Boilerplate
No YAML/config overhead. Send JSON, get JSON. SDKs handle authentication, retries and and webhooks.
Flexible & Future-Proof
Start unstructured. Adopt schemas later. Add metadata (e.g., "priority": "high")—without breaking changes.
Observability Built-In
Track message flows with GET /messages?context_id=. Integrate Prometheus/Grafana via webhooks.
No Vendor Lock-In
Self-host or deploy anywhere. Export your data anytime—messages are stored as raw JSON.