All posts
Developer tracking7 min read

Engineering visibility for CTOs: a practical playbook to track developer work

For CTOs and heads of engineering: engineering visibility as six questions answered from merged commits, and a weekly, monthly and quarterly rhythm to track developer work from Claude.

The DeployIt Team

We build DeployIt, an MCP server that lets a whole company ask its code from Claude and track each developer's work.

Engineering visibility for CTOs comes down to six questions answered with facts: who shipped what, was it reviewed, was it tested, what broke afterwards, how good is the code, and what rests on one person. All six can be read from merged commits, which is how DeployIt lets you track developer work from Claude: one fact per merged commit, credited to its author.

This playbook is for a CTO or a head of engineering who can no longer read every pull request. It has four steps: tell the team, connect the repositories, ask on a fixed rhythm, read each figure by its rule.

In your Claude · example

Give me last month, developer by developer.

Here is last month for the three tracked developers. Each score reads with the number of commits on its row.

DeveloperMerged commitsReviewedTestedCode score
Camille R.4671%52%3.9
Idriss B.3860%31%3.2
Léa M.3566%40%4.1
Drawn from your commits119 merged commits · sample data

What engineering visibility for CTOs means: six questions

"Visibility" is too vague to manage with. Replace it with questions that have an answer in the repository.

QuestionThe fact that answers itHow to read it
Who shipped what?Commits merged into the tracked branch of each repository, with their author, pull request and moduleCounted when they arrive; bots excluded
Was it reviewed?Pull requests reviewed by a colleagueNot the author, not a bot
Was it tested?Changes that came with testsA share of the merged changes
What broke afterwards?Defects: the commit that a fix came to repair"Confirmed" or "to confirm"; the last 60 days are not mature
How good is the code?A code score from 1 to 5 on five criteria, set by a language modelAlways with its number of commits; an indication
What rests on one person?Who wrote each module over the last 12 monthsModule by module

These are facts about individuals. DeployIt tracks each developer, commit by commit, and says so plainly: that is what its statistics are for.

Step 1: tell the team before you connect anything

The terms of service set the framework. The company informs the tracked people before it connects its repositories. A score set by a model is an indication, never on its own the reason for a decision. The human decision prevails over a defect counted by the machine.

Three things to say to the team:

  • What is tracked: merged commits, review by a colleague, tests, defects, and a code score set by a model.
  • Who reads it: the administrators, the people they give the Leadership profile to, and any data tool the company connects with a key.
  • What a person can ask for: an export of everything DeployIt knows about them, and its erasure.

During the trial or on a plan, tracking begins the moment the repositories are connected, and reads 6 months of history. The statistics are on by default; an administrator can remove a person from tracking, or turn them off. The privacy page says what a tracked developer can expect.

Step 2: connect the repositories, not a new pipeline

You do not need to join Git, CI and ticket events in a warehouse to answer the six questions. Connect the repositories you choose, from GitHub or GitLab. From then on, each merged commit becomes a fact, written once and updated every night, and every row links back to its commits.

Then give the Leadership profile to the people who should read each developer's figures. It opens the statistics of the team and of every developer, without code or file paths (filtered and tested; no filter is infallible). Example: the CTO and two engineering managers.

If the same figures must sit in your warehouse next to incidents and costs, a Data tool key reads them in bulk: see developer metrics in your data warehouse.

Step 3: track developer work on a fixed rhythm

Three requests, three rhythms.

RhythmWhat you ask ClaudeWhat comes back
Every week"What changed this week?" (the what_changed prompt)What landed on the tracked branch, grouped into deliveries, without naming anyone
Every month"Give me last month, developer by developer." (the team_report prompt)The team's figures and one row per tracked developer
Every quarter"Build this developer's record for the quarter." (the developer_profile prompt), then "Which modules depend on a single person?"One developer's record over the period; who wrote each module over the last 12 months

The prompts bring back the same table from one period to the next. That is what lets you read a trend.

For anything else, ask in your own words. Claude reads what exists with stats_schema, then sends a structured query with stats_query: a dataset, a period, filters, groupings, measures. It is never free-form SQL. Example: "Who introduced the defects we fixed this quarter?"

The weekly question is open to the Everyone, Developers and Leadership profiles, so product and support read the same deliveries as you. The quarterly record is the base of a 1:1: see engineering 1:1 with data. The monthly table is detailed in developer performance metrics from merged commits, and the last question in the bus factor of a codebase.

Step 4: read each figure by its rule

Arguments about engineering figures are usually arguments about definitions. Here the definitions are fixed: stats_schema returns the definition of each measure, and the time the statistics were last computed. Four rules are worth knowing by heart.

  • The code score is an average. It always comes with the number of commits it covers.
  • A defect shows up after the fact. The last 60 days are not mature, and the count can still rise. A defect stays "to confirm" until an administrator or an automatic investigation confirms it; only an administrator dismisses it.
  • A closed period is not frozen. A late commit, two identities merged or a defect dismissed can change it, and every summary states when it was last computed.
  • Only work that reaches the tracked branch counts, when it arrives. Bots are excluded, detected coding agents are counted separately, and an undeclared agent counts as human work.

Example, from the table at the top: Idriss B. shows 3.2 over 38 commits, with 31% of his changes tested. That is a question for his next 1:1 (which changes went in without tests, and why?), not a conclusion about him.

What this playbook does not show

  • Deployments and incidents. The five software delivery metrics of DORA (change lead time, deployment frequency, failed deployment recovery time, change fail rate, deployment rework rate) are all defined on deployments, according to dora.dev, checked on 17 November 2026. DeployIt reads what reaches the tracked branch and does not compute them. Keep them in your delivery pipeline and your incident tooling.
  • Work that has not landed. A branch still open, a prototype, a design document.
  • What customers do with a release, and how the team feels. Ask them.
  • A screen to watch. DeployIt has no dashboard and emails no report. The table exists when you ask Claude for it.

For a board, the same facts fit on one slide: see how to explain engineering velocity to investors. The product as a whole is described in DeployIt: ask your code and follow each developer's work from Claude, and the datasets in the statistics section of the docs.

What DeployIt does here

DeployIt is an MCP server. Your company gets one link, https://mcp.deployit.ai/<company>/<code>; each person adds it to Claude (on the web, on the desktop or in Claude Code) and signs up with a name and an email, which is not verified; their questions are logged under that name. Everyone starts with the Everyone profile: questions about the product, answered from the code; a filter removes the code, the file paths and the names of team members it recognises from the answers, without proving their absence, and a leak test tells you what still gets through. An administrator gives the Developers profile (excerpts) or the Leadership profile (developer statistics) to whoever needs it. Repositories are connected from GitHub or GitLab; DeployIt makes no commit, branch or pull request, and only adds one webhook per repository.

A working copy of the code is kept while a repository is connected, hosted by DigitalOcean in London (UK); at indexing, each source file is sent to the model to be summarised, then the passages needed for an answer and the diffs of the commits that are scored; OpenAI's models by default, or the provider whose key you bring.

Next step: open the trial, connect one repository, add your company link to Claude by the three steps of Connecting, and ask for last month, developer by developer (the table needs the Leadership profile, given by an administrator; the demo carries no statistics, so run this on your own repositories).

Price. Free plan: 15 questions a week for the whole company. A 30-day trial, no card. Plans from €290 a month excluding tax for up to 25 tracked developers: see pricing.

Ask your code from Claude

30-day trial, no card. Or try the demo first: add https://mcp.deployit.ai/demo to your Claude connectors.

Connect my code

Frequently asked questions

What is engineering visibility for a CTO?

It is the ability to answer, with facts, who shipped what, whether it was reviewed and tested, what broke afterwards and which modules rest on one person. Those answers are in the merged commits of your repositories. DeployIt writes one fact per merged commit, and a person with the Leadership profile asks Claude for the table.

How can a CTO track developer work from Git?

Count what reaches the tracked branch of each repository, author by author: merged commits, pull requests reviewed by a colleague, changes that came with tests, and the commits a later fix had to repair. DeployIt does this for each tracked developer and adds a code score from 1 to 5, always read with its number of commits. Claude composes the report by month, quarter, half-year or year.

Do you have to tell developers their work is tracked?

Yes. DeployIt tracks each developer, commit by commit, and its terms of service require the company to inform the tracked people before it connects its repositories. Tracking starts when the repositories are connected, during the trial or on a plan, with 6 months of history; an administrator can remove a person or turn the statistics off. A score set by a model is an indication, never on its own the reason for a decision.

Which engineering figures should a CTO look at every month?

Five, for the team and for each developer: merged commits, the share of pull requests reviewed by a colleague, the share of changes with tests, defects confirmed or to confirm, and the code score with its number of commits. Read them against the previous month, and remember that defects from the last 60 days are not mature. Once a quarter, add the modules that depend on a single person.

Does DeployIt compute DORA metrics?

No. The five software delivery metrics listed on dora.dev (change lead time, deployment frequency, failed deployment recovery time, change fail rate and deployment rework rate; checked on 17 November 2026) are all defined on deployments. DeployIt reads what reaches the tracked branch of each repository and says who shipped, reviewed, tested and fixed it. Keep deployment and incident measures in your delivery pipeline.

Continue reading

Connect my code