mentat

support and sales ping engineering all day. why is a senior engineer answering “how does this work?”

you onboarded a dozen new hires. why does ramping them up still take months?

you hired 40 engineers. why aren't you delivering faster?

Different teams, one cause: the answers live in a few people's heads, and getting them means interrupting those people.
Mentat answers what your team has already written, and reconstructs the reasoning they never wrote down.

create a workspace →

$49.99 per seat per month, 1 to 200 seats, monthly or annual

one question, four people paying for it

who
what happens
cost
new hire
Asks "how does auth work?" and gets sent to a stale wiki. Follows it, breaks their setup, spends 3 hours reverting. Never asks again.
Ramp time doubles. Nobody knows they're stuck.
senior eng
Said "check the wiki" mid-flow because they couldn't stop. Rational, but the wiki is stale and they know it.
Deep-work hours go to re-explaining, not building.
eng manager
Velocity dipped after adding 3 engineers and can't explain why.
Headcount ROI questioned by CTO.
cto
"We hired 40 people. Why did Q2 launch deadline slip?" Knows the answer but has no data to frame it.
Absorbs the blame at the board meeting.
3.5–4.5h daily focus time for the median engineering team, barely half an 8-hour day Worklytics, 2025
58% of developer time goes to understanding existing code, not writing new code Xia et al., 2018
8h/wk each engineer loses to fixable friction and poor documentation Atlassian + DX, 2024

works with the tools your team already uses

GitHubgithub
Azure DevOpsazure devops
Linearlinear
Slackslack
Microsoft Teamsms teams
Buzzbuzz
Confluenceconfluence
Notionnotion
Freshdeskfreshdesk

plus an MCP server, so Cursor and your own agents pull the reconstructed reasoning directly.

answer. detect. reconstruct. route.

Senior engineers won't write docs. Mentat asks the right questions instead.

01 answer

answer what's written

Mentat ingests code, GitHub, Slack, Notion, and Confluence, and answers questions against all of it, plus the reasoning it reconstructs.

  • code
  • github
  • slack
  • notion
  • confluence
02 detect

find the gap

Mentat finds what's missing, stale, or never written, and identifies who holds the live answer, and who else is close.

Mentat sources view showing GitHub repos, Slack channels, Notion pages, and document counts after connecting an org's tools

Channels, commits, PRs, issues, pages. Every gap mapped to who holds the answer.

03 reconstruct

reconstruct the reasoning

A manager dispatches a ~20-min async interview. Mentat writes the questions, conducts it, and captures the reasoning that existed in no document before.

A live Mentat interview capturing a senior engineer's reasoning about payment processing architecture

Live interview. The follow-ups that surface the why.

04 route

route it

The whole team and their coding agents pull the reasoning through the MCP server, against the work in front of them.

Mentat knowledge map dashboard showing where engineering expertise concentrates across product areas

Knowledge map. Self-serve with attribution.

Every reconstruction makes the answers better. The loop runs continuously.

every other tool starts from what someone wrote

cursor / copilot writes code
coderabbit reviews diffs
glean / guru searches and acts on what was recorded
notion ai / confluence ai summarizes existing pages
internal rag retrieves what was indexed
mentat reconstructs the reasoning no document holds

Every one of them answers from something that was already recorded. Mentat finds what was never recorded, asks the person who knows, and writes the answer back where the next person will hit it.

and it doesn't stop at mentat. your ai tools query the reconstructed reasoning directly over mcp, so cursor and your own agents pull the why at the moment they need it.

Claudeclaude Cursorcursor

latest from mentat

resources

whitepaper, web + pdf security whitepaper how the single-tenant vpc deployment works, every egress destination enumerated, and why phone-home is boot-enforced off. read → one-pager, pdf architecture overview how mentat reconstructs and serves the reasoning behind decisions. data model, knowledge construction, retrieval, on a single page. download pdf → one-pager, pdf product overview what mentat is, who it's for, and the workflow it changes. the one-page product summary. download pdf →

alexander milovidov

Portrait of Alexander Milovidov

founder & product engineer, integrity evolution inc. chapel hill, nc

Alexander started as an engineer at IBM and moved into product leadership. Samsung, Nike and TIBCO each launched a global platform he led, and he also led product at SAS.

One Samsung hiring push brought in 16 product managers in two months, and his onboarding playbook cut their ramp from two months to two weeks. Since January 2026 he has written Mentat's code himself.

see the reasoning behind your architecture that no document holds. start with one repo.

create a workspace →