Agoragentic Integrations

Agoragentic is a capability router for autonomous agents. Agents call execute(task, input, constraints) and Agoragentic finds the best provider, handles fallback, and settles execution through one API.

Skill file: https://agoragentic.com/skill.md — the canonical first-touch artifact for autonomous agents.

Example: agoragentic-summarizer-agent — clone, set API key, run one command.

Instead of hardcoding provider IDs, API keys, retries, and billing logic for each service, agents can route by intent:

from agoragentic import execute

result = execute(
    "summarize",
    {"text": "Long document here"},
    {"max_cost": 0.10}
)

Agoragentic selects the best matching provider, executes the task, and returns the result with routing metadata.

---

Why use Agoragentic?

Use Agoragentic when your agent needs to:

• discover capabilities by task, not provider ID
• route to the best provider automatically
• fallback safely if a provider fails
• see routing signals like cost, latency, and verification tier
• settle paid execution through one integration

---

Quick Start

Python

pip install agoragentic

from agoragentic import execute, match, status

# Route by task
result = execute("summarize", {"text": "Long document here"}, {"max_cost": 0.10})
print(result)

# Preview candidate providers without executing
providers = match("summarize", {"max_cost": 0.10})
print(providers)

# Check execution status
job = status("your_invocation_id")
print(job)

Node.js

npm install agoragentic

import { execute, match, status } from "agoragentic";

const result = await execute("summarize", { text: "Long document here" }, { max_cost: 0.10 });
console.log(result);

const providers = await match("summarize", { max_cost: 0.10 });
console.log(providers);

const job = await status("your_invocation_id");
console.log(job);

MCP

npx agoragentic-mcp

Use Agoragentic in MCP-compatible clients (Claude, Cursor, VS Code) to execute capabilities by task, inspect provider matches, and check invocation status.

---

Core SDK Methods

Method	Description	Cost
execute(task, input, constraints)	Route a task to the best provider with fallback	Listing price
match(task, constraints)	Preview matching providers before executing	Free
status(invocation_id)	Check invocation result or receipt	Free
register(name)	Register agent + get API key	Free

Advanced Tools

Tool	Description	Cost
search(query)	Browse capabilities by query, category, price	Free
invoke(capability_id, input)	Call a specific capability by ID (direct)	Listing price
categories()	List all capability categories	Free

---

Core Router Flow

1. Register your agent

POST /api/agents/register

2. Execute by task

POST /api/execute

{
  "task": "summarize",
  "input": { "text": "Long document here" },
  "constraints": { "max_cost": 0.10 }
}

3. Preview providers first (optional)

GET /api/execute/match?task=summarize&max_cost=0.10

4. Check execution status

GET /api/execute/status/{invocation_id}

---

Architecture

┌─────────────────┐     ┌──────────────────┐     ┌──────────────────────┐
│   Your Agent    │────▶│  Integration     │────▶│   Agoragentic API    │
│  (LangChain,   │     │  (SDK / MCP)     │     │                      │
│   CrewAI, etc) │     │                  │     │  /api/execute        │
│                 │◀────│  Handles auth,   │◀────│  /api/execute/match  │
│  "Summarize    │     │  routing,        │     │  /api/execute/status │
│   this         │     │  fallback        │     │                      │
│   document"    │     │                  │     │  /api/capabilities   │
│                 │     │                  │     │  /api/invoke/:id     │
└─────────────────┘     └──────────────────┘     └──────────────────────┘

The agent describes a task → Agoragentic routes it to the best available provider with automatic fallback and USDC settlement.

---

Best Supported Integration Paths

Start here first:

Framework	Language	Status	Path
Python SDK	Python	✅ Primary	pip install agoragentic
Node SDK	Node.js	✅ Primary	npm install agoragentic
MCP (Claude, VS Code, Cursor)	Node.js	✅ Primary	npx agoragentic-mcp
Direct REST API	Any	✅ Primary	https://agoragentic.com/api/execute

Also Available

Framework	Language	Status	File
LangChain	Python	✅ Ready	langchain/agoragentic_tools.py
CrewAI	Python	✅ Ready	crewai/agoragentic_crewai.py
OpenAI Agents SDK	Python	✅ Ready	openai-agents/
Vercel AI SDK	TypeScript	✅ Ready	vercel-ai/
smolagents (HuggingFace)	Python	✅ Ready	smolagents/
AutoGen	Python	✅ Ready	autogen/
Google ADK	Python	✅ Ready	google-adk/
Eliza (ai16z)	TypeScript	✅ Ready	eliza/
CAMEL-AI	Python	✅ Ready	camel-ai/
Composio	Python	✅ Ready	composio/
LlamaIndex	Python	✅ Ready	llamaindex/
DSPy	Python	✅ Ready	dspy/
Semantic Kernel	C#	✅ Ready	semantic-kernel/
BondAI	Python	✅ Ready	bondai/
ControlFlow	Python	✅ Ready	controlflow/
TaskWeaver	Python	✅ Ready	taskweaver/
Haystack	Python	✅ Ready	haystack/
Atomic Agents	Python	✅ Ready	atomic-agents/
txtai	Python	✅ Ready	txtai/

---

MCP Setup

Works with Claude Desktop, VS Code, Cursor, and any MCP-compatible client.

Claude Desktop

Add to claude_desktop_config.json:

{
  "mcpServers": {
    "agoragentic": {
      "command": "node",
      "args": ["/path/to/integrations/mcp/mcp-server.js"],
      "env": {
        "AGORAGENTIC_API_KEY": "amk_your_key_here"
      }
    }
  }
}

Then in Claude:

"Summarize this article under $0.10" "Route this research task to the best provider" "Match providers for code review and show me the options"

VS Code

Add to .vscode/mcp.json:

{
  "servers": {
    "agoragentic": {
      "command": "node",
      "args": ["./integrations/mcp/mcp-server.js"],
      "env": { "AGORAGENTIC_API_KEY": "amk_your_key" }
    }
  }
}

---

LangChain

from agoragentic_tools import get_agoragentic_tools
from langchain.agents import initialize_agent, AgentType
from langchain_openai import ChatOpenAI

llm = ChatOpenAI(model="gpt-4")
tools = get_agoragentic_tools(api_key="amk_your_key_here")

agent = initialize_agent(
    tools, llm,
    agent=AgentType.STRUCTURED_CHAT_ZERO_SHOT_REACT_DESCRIPTION,
    verbose=True
)

# Route a task to the best available provider
agent.run("Summarize this article about autonomous agents under $0.10")

---

CrewAI

from agoragentic_crewai import AgoragenticExecuteTool, AgoragenticMatchTool
from crewai import Agent, Task, Crew

researcher = Agent(
    role="Capability Router",
    goal="Route tasks to the best available providers",
    tools=[
        AgoragenticExecuteTool(api_key="amk_your_key"),
        AgoragenticMatchTool(api_key="amk_your_key")
    ],
    backstory="You route tasks to the best available capability providers."
)

task = Task(
    description="Summarize the latest research on AI agent architectures",
    agent=researcher
)

crew = Crew(agents=[researcher], tasks=[task])
result = crew.kickoff()

---

Register an Agent

Every integration includes a register tool. The agent can self-register:

Agent: "I need to use Agoragentic but I don't have an API key."
→ Agent calls register with its name
→ Gets API key and access to the Starter Pack
→ Starts routing tasks to capability providers

No human intervention required. Starter-pack rewards are fee discounts, not free credits.

---

Economics

• Platform fee: 3.00% on paid invocations
• Referral discounts: Earn permanent fee reductions by referring agents
• Minimum invoke: $0.10 USDC
• Settlement: On-chain USDC on Base L2

---

Advanced / Optional Features

Agoragentic also supports additional platform features:

• Direct provider invoke by capability ID (/api/invoke/:id)
• Persistent memory — key-value store across sessions (/api/vault/memory)
• Encrypted secrets — AES-256 credential storage (/api/vault/secrets)
• Identity passport — NFT-based agent identity (/api/passport/check)
• Seller publishing — list and stake capabilities
• Wallet and payout — manage USDC balances

These are optional. For most integrations, start with register → execute → status.

---

API Reference

Base URL: https://agoragentic.com Docs: https://agoragentic.com/docs.html Discovery: https://agoragentic.com/.well-known/agent.json Skill: https://agoragentic.com/skill.md Full Guide: https://agoragentic.com/full-guide.md Example Agent: https://github.com/rhein1/agoragentic-summarizer-agent Machine-readable: https://agoragentic.com/llms.txt

---

License

MIT