📄️ Agents Overview
Fyso is a backend-as-a-service that AI agents can operate directly. An agent can create data models, store and query records, enforce business logic, manage users, and deploy sites — all through tool calls or HTTP requests. No manual setup required: connect, select a tenant, and start building.
📄️ Authentication
Every request to Fyso needs two things: proof of identity and a tenant context. This page covers all authentication methods with working examples.
📄️ Bot Identity
Bots are service accounts that authenticate against the Entity API using a name and a long-lived secret. They receive short-lived JWTs (1 hour TTL by default) and carry scoped permissions defined at registration time.
📄️ MCP Quickstart
A step-by-step walkthrough of the core Fyso MCP workflow. Every tool call here is real -- you can copy-paste these into your MCP client and they will work.
📄️ Building a Complete App
Build a customer support ticket system from zero to deployed, with users, roles, business rules, views, API access, and a knowledge base. Every tool call is real.
📄️ Users & RBAC
Every tenant has its own user table, completely isolated from other tenants. Users authenticate with email + password and get a JWT. Roles control what they can do.
📄️ Channel Tools
Channels let a tenant expose custom MCP tools that other tenants and agents can discover and execute. Think of it as publishing an API built from your entity data, without writing backend code.
📄️ Web Widget
The web widget lets you embed a Fyso agent as a floating chat bubble on any website. Add one `` tag and the widget handles everything: session management, message delivery, and UI rendering.
📄️ Agent Memory
Agent memory lets an agent retain facts about an end user across separate conversations. When enabled, facts are extracted from the conversation after it ends and injected back into the system prompt on subsequent turns with the same user.
📄️ Embeddings & Semantic Search
Fyso provides two embedding systems that enable semantic (meaning-based) search across your data:
📄️ Agent Messaging
Fyso agents can send and receive messages to each other using the agentmessages system entity. This enables multi-agent pipelines where agents coordinate work asynchronously.
📄️ Requisitos de modelo para tool calling
Los agentes de Fyso usan un loop de tool calling: el modelo recibe una lista de herramientas (queryrecords, createrecord, updaterecord), las llama para obtener o modificar datos, y luego devuelve una respuesta de texto. Este loop depende completamente de que el modelo pueda emitir respuestas estructuradas con toolcalls.
📄️ RGPD / GDPR Compliance
Fyso provides built-in compliance primitives for operating AI agents under GDPR (called RGPD in Spanish/French contexts). This covers three areas: Data Processing Agreement acceptance by the tenant admin, session-level AI consent from end users, and the right to erasure via data suppression.
📄️ Auth for Static Sites
This guide covers authentication strategies for static sites deployed on Fyso. Choose the method that fits your use case.
📄️ External Agent Identity
External agents — processes running outside Fyso, such as Claude Code agents or autonomous scripts — can register a persistent identity with a tenant. This identity lets them receive messages via SSE without polling and participate in multi-agent pipelines.
📄️ MCP vs REST API
Fyso offers two integration paths: MCP (Model Context Protocol) for AI agents, and a REST API for traditional HTTP clients. This guide compares both approaches.
📄️ Agent Test Panel
The test panel lets you chat with an agent directly from the admin panel and inspect every execution detail without deploying to a channel first.