docenta ingests the code, every commit you authored, and every Claude Code and ChatGPT session, and serves them to your agent over MCP with citations and a token budget. "When did we decide that, and why" is answerable again.
An illustration of the shape of an answer. The files are made up; the behavior is docenta's.
On 227 judged questions over a real machine, the semantic tool put the truth in the top ten for 114; ripgrep managed 44 and read 2.7 MB per question to get there against 20 KB.
A repository map, co-change history and claims per module. The path lane knows a filename-shaped question when it sees one. Sessions and commits are searchable the same way.
content_search, semantic_search, evidence_pack, entity_card and timelines over MCP. Every page discloses what was searched, what was dropped and how many documents held every term.
Numbers from docs/eval in the docenta repository, reproduced by the head-to-head rig. Nothing here is a benchmark score of a model.
Back to the collection · also for lawyers, insurance, home, mail, developers, researchers
docenta is a commercial product of SKY, LLC, now in private preview. Tell us what your agent needs to remember and we will tell you when the door opens.
request early access or check back here: the public beta will be announced on docenta.ai