Developer platform

The full agent stack, already assembled.

Grounding, rules, tools, routines, actions, and every conversation surface run on one open-source platform. Build the agent your product needs instead of rebuilding its infrastructure.

Run it locally Read the docs

one deployment

Same agent, every surface

Knowledge + directives + routines
Radioso runtime
Web
API
SDK
MCP

Every feature open source

Nothing gated behind a paid tier

Bring your own LLM

Your keys, no inference markup

Your data stays yours

Postgres you can export — or host

Cloud or self-hosted

Same platform, your call

Inside Radioso.

Assistant

one turn loop · steer, act, follow through

Directives

steer
confirm before acting
stay concise
+ your rules

Standing condition → action rules. Shape behavior, no new capability.

Skills

act

retrieval.answer

ground answers in your docs

Your own

register a capability

retrieval draws on your docs → Postgres + pgvector

Routines

guide

Contact a human

collect → confirm → send

Your own

author a routine

Stateful multi-step routines, resumed across turns.

LLM
Grounded answer + citations·async actions (outbox)
Web chatPublic chatWebsite embedREST APITypeScript SDKMCP server

One agent. Use it anywhere.

Every interface reaches the same runtime and knowledge. Add a ready-made surface or build your own without creating another copy of the agent behind it.

Browse the API reference

Website embed

one script tag

Put a hosted conversation in your product or marketing site. Origin policy, appearance, and the agent behind it stay under your control.

REST API

plain HTTP

Drive conversations and product workflows from any stack without adopting a framework-specific runtime.

TypeScript SDK

typed client

Use the same API with types and first-class primitives for applications already built in TypeScript.

MCP server

@radioso/mcp-server

Expose agent conversations and workspace documents to MCP clients, with the same retrieval, rules, and citations attached.

Working assistant in under five minutes.

The commands download Radioso and open the project directory. The bootstrap script then prompts for your AI provider credentials, generates secrets, and starts the full stack with Docker Compose. Register, upload a document, ask a question.

Read the docs Setup guide
~/radioso
curl -fsSL https://github.com/radioso-ai/radioso/archive/refs/heads/main.tar.gz | tar -xz
cd radioso-main
# requires Node 24 + Docker Desktop
./run-dev.sh

# then open
# http://localhost:3000   web app
# http://localhost:8080   API

Start with a working platform, then make it yours.

Run Radioso locally, inspect the source, and follow the docs from your first document to an agent in production.

Read the docs View on GitHub