openAgent Documentation

openAgent is a platform for creating AI agents with persistent memory, auto-generated skills, and a public REST API. Create an agent, give it a name and description, and it gets its own API endpoint that anyone can call — no authentication required.

Auto-generated Skills

AI creates SKILL.md from your description

Persistent Memory

Agents remember across conversations

Public REST API

Every agent gets its own endpoint

Getting Started

1

Create an Account

Register at openagent.lol/register. You'll get access to the dashboard immediately.

2

Create an Agent

Give your agent a name and description. The platform generates a comprehensive SKILL.md that defines its behavior, tone, and capabilities.

3

Use the API

Your agent is instantly available at https://openagent.lol/api/v1/{slug}. Each agent gets a unique URL-friendly slug (e.g. my-agent) generated from its name. Send a message, get a response. No API key needed.

Try it now

curl -X POST https://openagent.lol/api/v1/my-agent \
  -H "Content-Type: application/json" \
  -d '{"message":"Hello, what can you do?"}'

Replace my-agent with any agent's slug.

How Agents Work

Each agent is a self-contained AI entity with its own personality, skills, memory, and knowledge base. When someone calls the API, the agent loads all its context, generates a response, and then autonomously updates its own memory and knowledge based on the conversation.

What happens when the API is called

1. Load SKILL.md (capabilities & personality)
2. Load MEMORY.md (learned from past conversations)
3. Load Knowledge Base (structured reference data)
4. Generate response (using all context)
5. Auto-update Memory (background)
6. Auto-update Knowledge (background)

Steps 5 and 6 happen asynchronously after the response is sent. The agent learns from every interaction.

Developer API

Every agent gets a REST endpoint at:

https://openagent.lol/api/v1/{slug}

Each agent gets a unique URL-friendly slug (e.g. my-agent) generated from its name. No API key, no authentication, no setup. Just send a POST request with a message and get a response.

GET /api/v1/{slug}

Returns basic agent info (name, slug, description, avatar, status).

POST /api/v1/{slug}

Send a message to the agent and get a response. Memory and knowledge are automatically updated.

Simple message
{ "message": "Hello, what can you do?" }
Multi-turn conversation
{
  "messages": [
    { "role": "user", "content": "What is 2 + 2?" },
    { "role": "assistant", "content": "2 + 2 equals 4." },
    { "role": "user", "content": "And if you multiply that by 3?" }
  ]
}
Response
{
  "response": "Hello! I'm here to help you with...",
  "agent": { "id": "abc123", "name": "My Agent", "slug": "my-agent" }
}

Options

"stream": true — SSE streaming response instead of JSON

"message" — Single message string (simple)

"messages" — Array of messages (multi-turn)

Code Examples

JavaScript / TypeScript

const response = await fetch(
  'https://openagent.lol/api/v1/my-agent',
  {
    method: 'POST',
    headers: { 'Content-Type': 'application/json' },
    body: JSON.stringify({
      message: 'Hello, what can you help me with?'
    })
  }
);

const data = await response.json();
console.log(data.response);
// => "Hello! I can help you with..."

Python

import requests

response = requests.post(
    'https://openagent.lol/api/v1/my-agent',
    json={'message': 'What can you help me with?'}
)

data = response.json()
print(data['response'])

cURL

Simple message
curl -X POST https://openagent.lol/api/v1/my-agent \
  -H "Content-Type: application/json" \
  -d '{"message":"Hello, what can you do?"}'
Streaming
curl -X POST https://openagent.lol/api/v1/my-agent \
  -H "Content-Type: application/json" \
  -d '{"message":"Tell me a story","stream":true}'

Memory System

Each agent has a persistent memory (MEMORY.md) that automatically evolves from conversations. After every API call, an AI evaluator analyzes the dialogue and decides whether to update the agent's memory.

What gets remembered?

  • User preferences and patterns
  • Important facts and decisions
  • Corrections to the agent's behavior
  • Recurring topics and contexts

Memory is evaluated asynchronously after each response. You can also manually edit the memory via the agent's Overview tab in the dashboard.

Knowledge Base

The knowledge base stores structured reference material that the agent uses in responses. Unlike memory (which stores context about interactions), knowledge entries are factual reference documents that are auto-extracted from conversations.

Knowledge Categories

generaltechnicalguidelinesreferencefaq

Knowledge is auto-extracted after each API call and can also be managed manually via the Knowledge tab on each agent.

Models

openAgent uses OpenRouter as the LLM provider, giving access to a wide range of models. The platform includes automatic fallback — if one model is rate-limited, it tries the next available.

Fallback Chain

1. Configured model (agent-specific)
2. deepseek/deepseek-r1-0528:free
3. meta-llama/llama-3.3-70b-instruct:free
4. qwen/qwen3-coder:free
5. google/gemma-3-27b-it:free