Updated
trovex for platform and DevEx teams
When a platform or developer-experience team rolls AI agents out to other teams, the hard part is consistency: every team's agents read a different slice of the docs and answer differently. trovex gives the org one canonical source of truth its agents read and write back to, served locally over MCP, so context is standard without a central service to operate. About 60% fewer tokens per lookup.
The problem a platform team owns
Your job is to make agents work the same way for every team, but each team points its agents at different docs and gets different answers. There is no shared notion of which doc is current, so behavior drifts team by team and you field the same questions about where the real source of truth lives.
One standard context layer, run locally
trovex serves the one current doc per query with a freshness marker, and each developer runs it against the repo locally. There is no central service for the platform team to operate and scale. The standard you ship is the index plus the freshness discipline, not a server you have to babysit, and any MCP client reads the same canonical answer.
How it rolls out
Make it the recommended setup in your developer environment: a one-line install, per-client config already documented under Setup. Adoption is bottom-up and standardized at the same time, with no procurement and no data leaving your machines.
When the rollout is the product
Standardizing agent context is the easy half. Making agents reliable across many teams, with scoped permissions, review gates, and context that stays current, is the operating problem platform teams own. trovex fixes the retrieval; tsukumo consults on the operating model, and trovex is the proof of how it is done.
FAQ
How do we standardize agent context across teams?
Point every team's agents at one canonical store. trovex serves the same current doc per query to any MCP client, so behavior is consistent across teams without each maintaining its own copy.
Does this add a service for the platform team to run?
No. trovex is local-first; each developer runs it against the repo. There is no central server to operate or scale, so you ship a standard, not an SRE burden.
How does it fit our existing dev setup?
It is an MCP server any client connects to (Claude Code, Cursor, Windsurf, Cline), configured per the /for setup pages. One install, no account, no data leaving your machines.
Ship one standard for agent context.
One canonical source of truth, run locally by each dev. No central service. About a minute to set up.
uv tool install trovex
Open source. No cloud, no API keys. Your docs never leave your machine.