Intelligence
is on.
Dhivon builds AI tooling for engineering teams. Our aim is to carry the reasoning behind a change from the first specification through implementation, review and the next team decision.
Our philosophy
“Intelligence is on” means a team can put what it knows to work at each stage of a change. A developer picking up a bug should be able to find the original decision, give an agent a clear task and see whether the proposed fix honours that decision.
Cadre captures the approved intent. MCP Gateway gives agents access to the tools a team permits. Our roadmap extends that work to choosing a model, writing the fix, reviewing the result and preserving what the team learns for the next developer.
Current products
Cadre
A spec-driven AI harness that turns an approved specification into a dependency-aware plan. Teams can review changes against the spec and resume interrupted work.
Explore Cadre ↗MCP Gateway
A self-hosted enterprise gateway for Model Context Protocol tools. Administrators set access, developers opt in, and the gateway checks policy when an agent discovers or calls a tool.
Explore MCP Gateway ↗Product roadmap
These four planned products address different moments in a team’s workflow. Their scope and release dates are still open.
LLM proxy / gateway
Give teams one place to choose which language models their tools can call. A team could use a faster model to summarise a bug report and a different one to reason through the fix.
Coding agent
Turn an approved task into a patch and tests. A developer could hand over a reproducible bug and get back a change they can run, inspect and revise.
Code review agent
Compare a proposed change with its specification and tests. Flag a missed requirement or an untested edge case while the author still has the reasoning fresh in mind.
Organisation knowledge base
Make past decisions available when a new question arises. The next engineer touching a workaround could find why it exists and what would make it safe to remove.