All your conversational agents. One self-hosted platform.
Steered by your rules, grounded in your knowledge, and reaching every surface — web, API, SDK, and MCP.
Ask RadiosoliveWhat is Radioso?
Radioso is a platform for self-hosted conversational agents — grounded in your data and following your rules1. An agent talks to your users, follows the procedures you author, and takes real action rather than just describing it1. It answers from your own content and cites what it used, so you can check it2. One deployment serves every surface: the web app, a REST API, a TypeScript SDK, a website embed, Slack, and MCP clients3. Your data and your model keys stay in your stack4.
Answers by RadiosoSelf-hosted
Runs on your own infrastructure
Bring your own LLM
Your keys, no inference markup
Your data stays put
Everything in your Postgres
Open standards
MCP, REST, and SDK included
It knows when to act — and when to ask.
Radioso diagnoses the problem, checks your rules, and does the work. When a rule says a person decides, it stops, asks, and hands over cleanly — with everything it already found attached.
When it genuinely needs a person, it hands off to a human — with the full conversation and every action it took attached.
That's the difference between talking about it and getting it done.
Inside Radioso.
one turn loop · steer, act, follow through
Directives
steerStanding condition → action rules. Shape behavior, no new capability.
Skills
actretrieval.answer
ground answers in your docs
Your own
register a capability
retrieval draws on your docs → Postgres + pgvector
Routines
guideContact a human
collect → confirm → send
Your own
author a routine
Stateful multi-step routines, resumed across turns.
Ask from any surface — the answer returns there.
Talks to your users and takes actions you need.
One platform, not five vendors.
A support agent from one vendor, cart recovery from another, lead qualification from a third — each with its own contract, its own integration, its own copy of your knowledge going quietly stale. Radioso runs them all in one place: your content once, your rules once, and a purpose-built agent on every surface.
five contracts · five integrations · five copies of your knowledge
Different teams, same engine.
Builders
Put an agent in your own product
Give your customers a grounded agent inside your app or site — drop in the website embed and public chat, or wire it in through the REST API, TypeScript SDK, or MCP server. No agent framework to adopt, no low-code canvas to drag boxes on. Bring your own LLM keys and self-host the whole thing.
- Website embed and public chat
- REST API, TypeScript SDK, MCP server
- Bring your own keys, self-hosted
High-volume CX teams
Resolve routine tickets end to end
Ground the agent in your help center and policies and it clears the routine queue itself — running routines like returns and escalations, honoring your directives every step. When a case is genuinely hard, it hands a person the full transcript with every action attached.
- Grounded in your help center and policies
- Runs routines like returns and escalations
- Human handoff with the full transcript
Knowledge-base teams
One knowledge base, cited answers everywhere
Ingest and curate your documents once, then let both sides of the house draw from the same grounded source. Your team asks in Slack or web chat; your customers ask through the embed or public chat — same citations, no second copy to keep in sync.
- Curate documents once, reuse everywhere
- Grounded answers with citations
- Team in Slack, customers in the embed
Working assistant in under five minutes.
The bootstrap script prompts for your AI provider credentials, generates secrets, and starts the full stack with Docker Compose. Register, upload a document, ask a question.
# requires Node 24 + Docker Desktop
./run-dev.sh
# then open
# http://localhost:3000 web app
# http://localhost:8080 API
Every feature, open source.
Every product feature is open source, and you can run the whole platform yourself. Enterprise Edition is only for multi-tenant deployments running Radioso at scale.
Open source
self-hostedEvery product feature ships in the open release — grounded answers, directives, routines, actions, every surface. Read the source, run the whole platform on your own infrastructure, and change what you need to.
Get the sourceEnterprise Edition
at scaleFor running Radioso across many workspaces and tenants at scale — multi-tenant, scaled deployments where one install serves a lot of people. Tell us what your deployment needs and we will figure out the shape of it together.
Contact usQuestions? Ask the product.
Short answers below — or press Ask the agent and watch the live agent up top answer it for real, grounded in the docs, citations included.
Every product feature ships in the open release. The Enterprise Edition covers multi-tenant deployments at scale.
Yes — the whole stack runs on your own infrastructure with Docker Compose. Documents, conversations, and vectors all stay in your own Postgres.
Bring your own keys — OpenAI, Anthropic, and more. You pay your provider directly, at their price.
It calls tools, fires webhooks, and runs multi-step routines — then hands off to a person with full context when it should.
Frameworks give you primitives to assemble yourself. Radioso ships the assembled product — ingestion, retrieval, agent runtime, chat UI, API, and MCP server, pre-wired.
Yes — a standalone MCP server ships with it. Your knowledge base becomes a tool any MCP client can call, citations attached.
Every organization deserves a voice.
For thirty years, using the internet meant clicking. Increasingly it will mean talking — to agents that book the flight, find the recipe, fill the order. Every organization will need a way to speak for itself in that conversation.
Radioso runs on your infrastructure, with your model keys, and your data stays in your Postgres. Your voice stays yours.