AI Agents

AI agents that work over your data, documents, and systems.

Not a chatbot bolted onto a website. A working assistant that queries your live data, answers from your real documents with citations, and takes controlled actions inside the systems you already use.

We build each agent on a platform we have already proven in real applications, so you get a polished, trustworthy assistant without starting from zero.

See what they can do

Assistant

Which suppliers are behind on deliveries this quarter, and what does our contract say about late penalties?

Reading supplier contracts

Querying delivery records

Preparing answer

Three suppliers are past their delivery windows this quarter. Your master agreement allows a 1.5% penalty per late week, capped at 10% of the order value.

supplier-agreement.pdf 94% deliveries (live)

Across your systems

One question, answered across every system.

Your information is scattered across a dozen different tools. A Radian agent reaches into all of them and hands back one clear answer, so people stop copying between tabs and just ask.

Ask what is going on with a client's account. The agent looks them up in your CRM, pulls their recent orders from the database, and scans Slack for the latest feedback. Then it replies with one summary and a link to every source.

The idea

An agent plans in language. It acts through code.

A language model is good at understanding what someone is asking and working out how to help. A Radian agent puts that to work through a defined set of tools, connected to your real data and the systems you already use. It acts only inside the limits you set, so you get something useful and stay in control of what it can do.

Grounded in your data

It answers from your documents and your database, not from guesswork, and every claim traces back to a source.

You stay in control

You decide what the agent can see and do. It works within the limits you set, so it helps with real work without overstepping.

Shows its work

You see what the agent is doing as it happens, and exactly where every answer came from.

Capabilities

One platform. Everything an agent needs to do real work.

We build each agent from the capabilities your work calls for. This is the range we draw from.

Ask your database a question

People ask a question and the agent queries your live database to answer. It uses RadianSQL under the hood, so access stays read-only and governed.

Answers grounded in your documents

The agent searches your real documents and answers from them, citing the exact source for every claim. Hybrid search combines meaning and keywords for accuracy.

Ingest documents and websites

Load PDFs, text, and Word-style files, including scanned or image-only PDFs read with vision. Point it at a website and pull that content into its knowledge too.

Connect to your systems

Through the open MCP standard, an agent can plug into the tools your team already works in, and new connectors slot in without rebuilding the agent.

HubSpot · Slack · CRMs · Help desks · File storage · Internal APIs

Take controlled actions

A defined set of actions the agent is allowed to take, such as looking up an order, drafting a proposal, or starting an email. You choose which actions it can perform.

Run the numbers and generate files

For heavier analysis the agent runs Python over query results, then produces downloadable Excel, PDF, or CSV files on demand. Gated for safety and off by default.

Coordinate specialists

A lead agent routes each question to the right specialist: a sales expert, a data expert, a CRM expert. Then it combines their work into a single answer.

The platform

Built on a platform, not from scratch.

We have already built the hard, reusable parts: the engine that orchestrates, retrieves, and connects, and the polished interface your team works in. Neither carries anything specific to one client, so your agent inherits all of it on day one. We add your data, your documents, your connectors, and your brand.

The engine

Proven engine, not a prototype

Orchestration across specialists, read-only database access, retrieval with citations, system connectors, and streaming. The part that makes an agent capable and safe.

Includes

Multi-agent routing · Retrieval · Read-only SQL · MCP connectors · Tools framework · Ingestion · Streaming

The interface

An interface people want to use

Streaming answers, live activity, citations, document preview, and cards, all reskinned to your brand. The part that decides whether people come back to it.