Colibri
Stop searching. Start asking.
BRAFT's flagship product. An agentic AI system that turns your internal documents into instant, sourced answers.
One question. Every document.
Without Colibri
- Information scattered across shared drives, mailboxes, and local folders with no unified search
- Team members spend 30+ minutes finding a single document, specification, or policy
- No way to leverage past work: each project, bid, or report starts from scratch
- When key people leave, their knowledge of decisions and precedents leaves with them
With Colibri
- One chat interface to query your entire knowledge base in plain language
- Sourced answers in seconds, with the document name, page number, and relevance score
- Automatic cross-referencing: Colibri surfaces relevant precedents and reusable content
- Institutional knowledge preserved and accessible to every authorized team member
The interface
One interface. Your entire knowledge base.
Colibri gives your team a single chat interface to query your entire document base. Upload files or connect your existing storage. Ask anything in plain language.
Every answer is traceable. Colibri cites the source documents and page numbers it relied on, and your team can open any reference to verify it directly. When the information isn't in your documents, Colibri tells you.
The tools panel lets users choose which sources to query: the internal knowledge base, connected databases, or the web.
“New tender: office HVAC renovation, 800m², Brussels. What relevant past projects do we have?”
HVAC Renovation — Avenue Louise, Brussels (2022)
Technical report · 3 sections matched · p. 14, 28, 41
Office Energy Audit — Ghent (2023)
Audit report · 2 sections matched · p. 7, 19
Energy Performance Assessment — Liège (2021)
Specification document · 1 section matched · p. 33
Illustrative example
Cross-referencing
Leverage every past project
When your team starts a new project, bid, or audit, Colibri doesn't just search for keywords. It identifies past projects with similar requirements, surfaces reusable content from previous deliverables, and flags relevant precedents across your entire document base.
A new tender comes in. Colibri finds three past projects with comparable specifications, pulls the relevant sections from their technical reports, and highlights where the current requirements differ. Work that used to take a full day of manual searching happens in a single query.
Under the hood
Not a chatbot. An agentic system.
Most AI document tools run a simple retrieval pipeline: search for similar text, paste it into a prompt, return the output. Colibri works differently.
When you ask a question, an Orchestrator agent decides which tools and data sources to involve. Specialized agents (a database expert, a document retriever, and others) fetch the relevant information. A Critic agent reviews the draft answer for accuracy and completeness before the Responder delivers the final output.
This architecture handles tasks that a simple search can't: generating a document from a company template, cross-referencing data across projects, or pulling structured data from a database and unstructured knowledge from PDFs in the same query.
The orchestration graph: agents collaborate to handle complex queries. The workflow is fully customizable.
Already in production
Deployed at Alteo Group across 12 companies
Colibri indexes over 50,000 documents for the Alteo Group, a Belgian group active in building energy services. Field engineers and project managers use it daily to find technical specs, compliance requirements, and past project data.
Data sovereignty
Hosted in Europe. Your data stays in Europe.
Colibri runs entirely on European infrastructure. Your documents are indexed and stored within the EU. Nothing is sent to non-European servers. GDPR-compliant by design, built for organizations where data sovereignty is a requirement.
See what Colibri can do for your team
Works with any document base, any industry. Let's talk about your use case.