ConsilienceOntology code intelligence
The same agent that reasoned over your knowledge now works on code. It builds a live map of your code and links it to the notes and concepts in your workspace, so it pinpoints what to fix with real context. It plans, finds, fixes, and runs it to verify.
Consilience gives the agent a living map of your code and bridges it to your workspace knowledge — so it works with context a text-search agent structurally can't reach.
It maps your code up front — symbols, calls, and imports — so it jumps straight to where something is defined and traces how it connects. It shines on meaning-based questions like “the function that compacts history when context overflows”, where the words you'd search for never even appear in the code.
CodeBridge ties your notes and ontology concepts to the code that implements them in one shared meaning-space. Ask by concept and it finds the code; rest on a function and it surfaces the design note behind it. All on your device.
* Code intelligence indexes your workspace on your device. The code map is built from your source structure — symbols, calls, and imports — and stays current as you edit, so it never guesses. Every change still goes through your approval.
The same plan → explore → run loop that reasoned over knowledge runs on code too.
The agent breaks the work into a checklist. It shows you what it will touch and in what order, and you can pause to review if you want.
It maps the structure of the code up front to pinpoint functions and how they call each other. It finds exactly what to fix fast, without reading whole files.
It shows every change at a glance and runs builds, tests, and scripts itself to confirm the result. Verified by actually running it, not by guessing.
It understands code function by function, pinpointing it as precisely and quickly as it once navigated your notes. It only reads what it needs, so it's more accurate with less effort.
Not just notes — it creates and edits any kind of file. Same agent, same workspace.
It runs builds, tests, lints, and scripts itself, reads the results, and decides its next move. It doesn't stop at writing code — it runs it and fixes it.
As the work progresses, each item is marked done in real time. You always see how far it's gotten and what's left.
Its train of thought, its messages, and even the edits landing in your documents all unfold in real time. You see exactly what it's doing as it happens.
From asking every time to running on its own, you set step by step how much the agent does for itself. Every change goes through your approval, just like your notes.
Local-first. The agent that reasoned over your knowledge plans, fixes, and runs code in the same workspace.