Fragmented notes
Knowledge is spread across files, chats and people; no one sees the whole picture.
What your organization knows,
in a single memory.
The knowledge layer behind every Butterfly application.
Butterfly Brain is the enterprise memory of the Butterfly platform: it captures decisions, definitions, rules and the context behind past work, organizes them and brings them back at the right moment. It is not a chatbot or an assistant; it is the foundation layer that other Butterfly applications read from and write to.
Runs on-premisesModel-agnosticHuman approval and audit trailWeb, iOS, MCP and REST
Why did we add a second approval step to purchase approvals?
DECISIONA second approval was added for requests above a set amountThe rationale and the team that made the decision are stored with the decision record.
DEFINITIONCritical purchase requestHow the organization uses this term and which rule it is tied to.
SUMMARYProcurement process meeting: decisions madeSummary of the conversation and next steps, linked to the related decisions.
Hybrid search: meaning and keywordResults by permission
Decision record
A second approval was added for requests above a set amount
- Rationale
- Why the decision was made, as recorded by the team that made it.
- Source
- TaskFlow requestMeeting summary
- Related
- Critical purchase requestApproval rule
- Access
- Authorized people and agents
Search and decision record · Ask in natural language; decisions, definitions and summaries come with their rationaleRepresentative Brain screen · sample data
ProblemKnowledge is scattered,
Knowledge is scattered,
context gets lost.
Decisions stay in chats, notes in files, context in people’s heads. As the team grows, the same questions get asked again and the same work gets rediscovered.
Recurring work
People say "we solved this before," but there’s no record, so the work starts from scratch.
Knowledge walking out the door
When someone leaves, the context leaves with them, and enterprise memory erodes.
How does it work?A working memory
A working memory
in three steps.
Brain keeps decisions and their rationale, the organization’s definitions and rules, and context from past work, and links them all in a knowledge graph.
- 1
Capture
It monitors work sessions and enterprise knowledge in the background, capturing information without manual entry.
- 2
Organize
AI removes duplicates, summarizes and tags decisions, so memory stays clean and current.
- 3
Recall
Ask in natural language; relevant records, decisions and summaries come to you ranked.
Trust and controlOn-premises, governed,
On-premises, governed,
model-independent.
Brain's value lies in holding only what your organization knows. That's why it is designed to run on your own infrastructure, isolated per organization and never shared with third parties.
Control and oversight
- Human review and approval before anything is published
- Model-agnostic: use the agents and models you approve
- Knowledge isolated per organization
- Immutable audit trail across applications
Deployment and data
- Runs on-premises or in your private cloud
- End-to-end encrypted channels for agent sessions
- Knowledge, code and prompts never leave your network
- Administration, permissions and auditing come from the Butterfly platform
How does it fit together?One layer.
One layer.
Many applications.
Brain sits beneath the applications. Each application reads context from Brain when it starts work and writes what it learned back to Brain when it finishes. Here’s how a piece of knowledge travels:
- A decision is made in TaskFlowA second approval is added to purchase requests above a set amount; the decision and its rationale are written to Brain.
- Brain adds the decision to institutional memoryThe decision is linked to the relevant definition and prior context, and becomes accessible to every authorized person and agent.
- Presentations Agent uses it in a presentationWhen the quarterly presentation is prepared, the new approval rule goes on the slide as the organization’s actual decision rather than generic text.
Butterfly is more than a low-code tool: application development, integration, process orchestration and AI agents run on a single BOAT platform, on top of the same knowledge layer.
ApplicationsBrain-powered
Presentations Agent
Coding Agents
App Marketplace and SDK
Brain-powered
applications.
Every Butterfly application is useful on its own; Brain makes it better. They all run on your own infrastructure, work with the AI models you approve and always keep a human in the loop.
Polished presentations from your content and data.
Point it at a document, a dataset or a project, and it prepares a structured, on-brand presentation draft. Slides reflect your organization’s definitions, past decisions and voice.
- Builds presentations from documents, data and decisions
- Content matches your company’s actual language
- You review and edit; nothing is shared without approval
From request to merge request.
Assign a request; the agent finds the right repository, loads company context from Brain, makes the change and opens a merge request. Manage sessions from the iOS, Android and macOS apps over an end-to-end encrypted channel.
- A human always gives the approval
- Monitor agent sessions from mobile and desktop
- End-to-end encrypted, model-agnostic
Build and distribute your own Butterfly applications.
The Butterfly SDK offers native clients in eleven languages, including Swift, Kotlin, Go and Python. App Marketplace distributes your internal apps with one-tap installation and over-the-air updates.
- Native clients in eleven languages, one shared SDK
- In-app feedback: a screenshot and note become a work item
- One-tap installation and over-the-air (OTA) updates
CapabilitiesCapabilities
Capabilities
out of the box.
Search by meaning and keywordHybrid retrieval: understands intent without missing exact word matches.
Decision and summary tracking"What did we decide, where were we?" answered in one place.
iOS appMemory in your pocket: search, read and share from mobile.
Access via MCP and RESTAccessible from inside the tools you already use, with no need to switch screens.
Automatic curationA scheduled AI worker merges duplicates and keeps memory well maintained.
Presentation generationA shareable slide deck is prepared from memory.
Which teams is it for?Teams that want to free institutional memory
Teams that want to free institutional memory
from dependence on individuals.
Knowledge-intensive, fast-growing teams
Managers, project and operations teamsWhy? ↻Why?
As the team grows, the same questions come up again; if the answers aren’t recorded, each new person gathers context from scratch.
What it doesBrings decisions, definitions and meeting outcomes together in one searchable memory.
Managers, project and operations teamsNew team members
Human resources and team leadsWhy? ↻Why?
If context lives in one person’s head, onboarding depends on that person’s time.
What it doesNew joiners find past decisions and the organization’s definitions without having to ask.
Human resources and team leadsSoftware and IT teams
Engineering and IT managementWhy? ↻Why?
If an agent writes code without knowing the organization’s rules, the review burden falls back on the team.
What it doesCoding agents work with company context, and every change is merged with human approval.
Engineering and IT managementNeeds assessment
Butterfly Brain and your organization: how good a fit?
8 questions, two minutes. At the end, see Brain’s fit scorecard for your organization and the reasons behind it.
- Scattered knowledge
- Corporate memory
- AI usage
- Data and control
Frequently askedQuick answers
Quick answers
about Brain.
Is Butterfly Brain a chatbot?
No. Brain is the enterprise knowledge layer that Butterfly applications read from and write to. It holds decisions, definitions and context; applications and agents use that context.
Do we enter information manually?
No need. Brain captures information from work sessions and enterprise knowledge in the background; AI removes duplicates and summarizes.
How do you search?
You ask in natural language. Hybrid retrieval, which combines semantic and keyword search, returns relevant records ranked by score. It is accessible via the web, the iOS app, MCP and REST.
Where does our data stay?
Brain runs on-premises or in your private cloud. Knowledge, code and prompts never leave your network and are not shared with third parties.
Which AI models does it work with?
Brain is model-agnostic; you can use the agents and models your organization approves.
How can we evaluate it?
The clearest path is a guided proof of concept (PoC) deployed on your infrastructure. Our team supports installation, configuration and migration from your existing tools.