Open Source · MIT

beebridge

Achieve real work objectives through AI — not just conversation.
Ontology graphs, minimal tokens, multi-model Waggle collaboration.

View on GitHub

Vision

beebridge exists so that people can achieve real work objectives through AI, not just have open-ended conversations.

To get there efficiently, the system takes two complementary approaches. First, an ontology-graph approach: tasks, districts, and bridges form an explicit graph of meaning and relationships, so the AI operates on structured knowledge rather than re-explaining everything from scratch. This keeps token usage to a minimum. Second, a goal-proximate history approach: context and memory are curated to stay close to the objective at hand, surfacing what actually advances the goal instead of accumulating irrelevant noise.

On top of that sits Waggle mode: a way to subscribe to and combine multiple AIs — local models, cloud providers, different tiers — and let them collaborate in conversation. The back-and-forth dialogue draws out more than any single model could deliver on its own, pushing past the inherent limits of whichever baseline you started with. The feel is deliberately human-like: people are not omniscient either, yet we still reach our goals by picking up tools, switching models, asking again, and iterating — and that is exactly the loop beebridge replicates.

Inside each bridge, work is embodied as bees — small, visible agents with concrete roles. A single objective grows organically into a hive: many bee-shaped steps, linked districts, and bridges, rather than one monolithic prompt. The bee metaphor is not decorative — it shapes how tasks are created, assigned, and connected.

The larger ambition is a network. As people use beebridge they produce workflows — graphs, exports, patterns — that they can share with each other. Over time, those shared workflows knit together into a broad AI-powered technology network: an interconnected mesh of practice that no single team could build alone. That collective growth — from individual hives to a shared ecosystem — is the direction beebridge is meant to evolve toward.

Core Concepts

Ontology Graph

Districts, bridges, and tasks form a navigable graph of structured meaning — not a linear chat log. Reuse context without re-explaining.

Minimal Tokens

Goal-proximate history keeps only what advances the objective. The graph eliminates redundant prompting, cutting token waste dramatically.

Waggle Mode

Like the bee waggle dance, models share knowledge with each other — filling gaps no single AI could cover. Subscribe to multiple AIs and let them collaborate beyond one model's limits.

Bees & Hives

Each task is a bee. One objective grows into a hive of connected steps across districts and bridges — visible, exportable, shareable.

Workflow Sharing

Export and import bridge-settings JSON. Share district graphs, task patterns, and bee configurations between teams and machines.

Network Ambition

Individual hives link into a broad AI-powered technology network — an interconnected mesh of practice that evolves collectively.

Architecture at a Glance

Gateway (Node)                    Extension (Flower)                 Chrome tab
  CDP relay WS  <--------------->  chrome.debugger.sendCommand  <-->  page
  + task loop        token            onEvent -> relay
  WebSocket /ws  (flower.register metadata only)

Full architecture details →

Guides

CLI Reference

Commands, env vars, daemon mode, onboarding.

Web Settings

Connection, auth profiles, model policy, workspace, diagnostics.

Bridge Graph

Districts, export/import, pipeline runs, upstream context.

Waggle Mode

Browser & API sub-modes, cost strategies, setup steps.

Quick Start

git clone https://github.com/newaibridge/beebridge.git
cd beebridge
npm install

npm run build --workspace=packages/shared
npm run build --workspace=packages/core
npm run build --workspace=@beebridge/gateway

# Start gateway
GATEWAY_TOKEN=dev-token npm run dev:gateway

# Start web UI (new terminal)
npm run dev:web

Then open http://localhost:3000, install the Chrome extension, and start creating districts and tasks.