One platform.
Three simple parts.
Cabinet is not an OKR tool, a coaching app, or a meeting assistant. It is the layer those things sit on top of: your files (strategy, OKRs, org chart, decisions, one-on-ones), a schedule that keeps them up to date, and four AI teammates that do the routine work for you.
Your files
Strategy, OKRs, org chart, decisions, one-on-one notes: all in plain text, in a folder you own, versioned by git.
The schedule
A simple list of recurring jobs: morning brief, weekly status, monthly review. Each one written down. Each one runs on time.
The AI teammates
Chief of Staff, Coach, Strategy Analyst, Risk Watcher. They read your files, write the briefs, surface the risks.
Your files
Your strategy as a folder. Yours forever.
Cabinet keeps your strategy as a folder of plain text files, connected by simple links, like a private wiki. The AI teammates read the same files you do.
# Q2 2026: Narrative
# Owners: [[org/sarah-kim]] · [[org/marcus-ren]]
# Source for: [[mbr/2026-05]]
## North star
Become the substrate every Fortune-class
company runs strategy execution on.
## Commitments
- [[okrs/emea]] · slipping ⚠
- [[okrs/product-velocity]] · on track
- [[okrs/cost-of-rollout]] · ahead
## Decisions log
- 2026-05-12 · pricing extension [[decisions/2026-05-12]]
- 2026-05-04 · hiring freeze partial [[decisions/2026-05-04]]
Versioned by git
Every change is a commit. Roll back a quarter, branch off a separate strategy, or compare two versions of a roadmap side by side.
No proprietary database
Cabinet stores nothing in a closed system. The same files open in Obsidian, Cursor, VS Code, or the Cabinet UI. Export is `cp -r`.
Linked like a wiki
A one-on-one file mentions [[okrs/emea]]. The AI follows the link to find the relevant OKR. Knowledge builds up the way you write it.
Searched by code, not embeddings
A real index, not a vector database. Reproducible, debuggable, no drift. The AI works on top of clean, structured retrieval.
The AI teammates
Four specialists. All reading the same files you do.
Each AI teammate has a single job, a clear permission scope, and access to a defined slice of your files. None of them sends your data to a general chatbot. Every action is grounded in files you can read yourself.
Chief of Staff
Runs the daily operation
Drafts the morning brief every weekday, the status every Friday, the monthly review brief on the first of the month. Reads your files, surfaces what is slipping, and drafts the follow-ups before you wake up.
Leadership Coach
Works alongside the manager
Sits in the one-on-one prep flow. Drafts kudos from real contributions. Helps frame the hard conversation before it happens. Never replaces the manager. Gives them what a great coach would give them.
Strategy Analyst
The numbers behind the deck
Compares OKR progress week by week. Builds the twelve-month metric grid for the monthly review. Calls out the movements that matter and ignores the noise that does not.
Risk Watcher
Looking out for what slips
Checks one-on-one logs, OKR progress, and dependencies every hour. Surfaces overdue follow-ups and unclear ownership before the next standing review.
The schedule
Every recurring task is a job written down.
Each scheduled task (morning brief, weekly status, monthly review brief) is a short, readable file. When it runs, what it reads, and what it writes. You can read it, change it, or version it like any other code.
# Daily 06:00 → today/index.md cron: "0 6 * * 1-5" agent: chief-of-staff sources: - okrs/ - 1on1s/ - decisions/ - mbr/ scope: read-only output: today/index.md prompt: | Read the OKR pages and 1:1 logs. Identify the top 3 actions for today. Surface any dropped commitments (1:1 promises past their due date). Draft replies to anything pending. Write the brief as the executive's Chief of Staff would, grounded in the actual files. Cite sources.
Every weekday · 06:00
Morning brief
Today's top 3, missed commitments, drafted replies, ready before you open your laptop.
Every Friday · 16:00
Weekly status + kudos
What moved this week, what is at risk, who deserves a callout.
1st of the month · 07:00
Monthly review brief
Twelve-month metric grid, scorecard, three things to discuss.
Quarter -1
OKR cycle
Drafting → reviewed → aligned → in progress → done.
Every hour, in the background
Risk check
Looks for things falling behind, missed commitments, and unclear ownership.
Integrations
Cabinet lives where work already happens.
We don't ask your team to log into another tool. The AI teammates post into Slack and Teams, pull from Linear and Jira, query Snowflake, and write back to your files, all running on your own AI provider keys.
Slack
Brief in DM
Microsoft Teams
Native channel agent
Reply drafts queued
Linear
OKR → issue mapping
Jira
Sprint roll-up
GitHub
PR-aware status
Notion
Two-way page sync
Snowflake
Live metrics in the brief
BigQuery
Bowler chart source
Postgres
Read replica
Anthropic
Your API key
OpenAI
Your API key
AWS Bedrock
Your AWS account
Azure OpenAI
Enterprise tenant
Plug in your own AI tools. No lock-in.
Cabinet exposes its data through MCP, an open standard. Your own AI tools (Claude Code, Cursor, anything that speaks MCP) can read your strategy files and add to them. Cabinet becomes the layer underneath, not the only window in.
- Open standard: no Cabinet-only SDK to learn
- Permissions per tool: read-only by default
- Every request and every change is logged
- Works with Claude Code, Cursor, Cline, and your own scripts
Your AI tools
Claude Code · Cursor · custom scripts
Cabinet open API
Scoped · audited · read-write
Your files
strategy/ · okrs/ · org/ · decisions/
Ninety minutes. Your strategy. A live demo on your org.
We set up Cabinet on your top three OKRs together, walk you through the morning brief, and leave you with a working folder you can use on Monday.