Ask AI · Engineering analytics and MCP

Ask in plain English.
Get the evidence back.

Ask questions about developer productivity, AI spend, code quality, and agent effectiveness using your organization’s Larridin data. Review the supporting evidence in Ask AI or connect through the Larridin MCP server.

Try Now Get Pricing

Ask about adoption, spend, velocity, quality or agent readiness View example answer

Trusted by AI-forward enterprises

Ask anything

One question, any part of the SDLC.

Spend, telemetry, agent traces, repo readiness: the same thread reaches all of it. Type the question the way the board will put it to you on Thursday, and follow up until the number holds.

Your data, not a generic benchmark · answers come from your own sessions, PRs, invoices and traces; comparisons use the scope and evidence returned by the selected analytics tool

Follow-ups keep the thread · ask "and only on Codex?" and the answer narrows without starting over

Every answer shows its working · the data pulled, the filters, the computation and the caveats, with the merged PRs and sessions it read, so the follow-up question is already answered

Written for the seat · the same question returns org-wide figures for an admin and a manager's own teams for the manager

How does AI-assisted output compare across teams?

Larridin AI

Sources: 1,204 PRs · GitHub · Jira · 12 weeks

AI-assisted engineers ship 1.6× the Engineering Output rate of human-only work on 3 of 4 teams. Growth is the exception: its lift is 7%, and its sessions most often close without a test run.

Team	AI-assisted	Human-only	Lift
Platform	4.1	2.6	+58%
Payments	3.8	2.5	+52%
Growth	2.9	2.7	+7%
Infra	3.6	2.2	+64%

Engineering Output per engineer per week, last 12 weeks. Same repos and scoring model.

What 3 things can I do to improve velocity?

Larridin AI

Illustrative example: 412 sessions in the selected period

Three changes, ranked by expected effect on your 12-week PR cycle:

Start every session with an explicit goal

Growth opens 61% of sessions without one. Sessions that state the goal take 31% fewer iterations.

Require a test run before the agent closes

Sessions that close without verification produce 2.7× more follow-up fixes. 38% of yours did last quarter.

Constrain the edit surface up front

Naming the files in scope yields 2.4× smaller diffs, which clear review faster.

MCP connector

Bring Larridin into your engineering assistant.

Larridin’s Beta MCP connector brings organization-scoped engineering analytics into supported assistants. OAuth identifies your account; enabled tools and analytics access depend on your organization and role.

1Add the server.In Claude Code, one command: claude mcp add --transport http larridin https://app.larridin.com/mcp/larridin. See the setup guide for Claude Desktop, Cursor, and VS Code.

2Sign in once.Run /mcp and complete the browser sign-in with your Larridin account. It is the same identity provider as the app, so no separate credential is issued.

3Ask.whoami confirms which tenant you are signed in as. From there query_larridin_analytics and ask_question answer in the chat.

Beta. The connector must be enabled for your organization. Review tool availability and access scopes with your administrator.

Claude larridin connected

What are we spending on AI coding tools this quarter, by source?

larridin query_larridin_analytics domain: ai_coding metric: cost_summary 12 weeks 0.8s

$48,200 net this quarter, up 24% on last. Most of the increase is Copilot on mobile since the July rollout. Codex and Cursor barely moved.

Source	Net	vs last quarter	Share
Claude Code	$20,400	+21%	42%
GitHub Copilot	$14,600	+56%	30%
Codex	$9,300	+6%	19%
Cursor	$3,900	+4%	8%

Sources: ai_coding/cost_summary, actual billed cost, Jun 13 to Sep 5, 4 sources. Org-wide view; a manager would see their own teams.

Ask about adoption, spend, velocity, quality or agent effectiveness

Works in

Claude CodeClaude DesktopCursorVS Code

Read the MCP setup guide

Ask AI Commands

Settings

Commands appear in the composer's slash menu for everyone in your organization.

Add Command

/velocity-dropExplain the change in Engineering Output per engineer per week over the last 4 weeks and name the biggest drivers

/one-on-one @employeePrepare a one-on-one briefing for

/deploy-freqHow often did we deploy last month compared with the month before, by service

/seat-reviewList seats with no session in the last 30 days, by tool, with the monthly cost of each

Add and Edit share one form. A command with an Employee parameter runs against the engineer tagged in the composer; the server adds the referenced employee after the prompt text at run time.

Custom commands

Turn the question you ask every week into a command.

Ask AI Commands live in Settings. Name the command, write the prompt text it should run, and it shows up in the slash menu for the whole organization. Add an @mention parameter and the same command runs about the engineer you tag.

No template to maintain · the author writes plain prompt text; there are no tokens to place, and the referenced engineer is inserted at run time

A briefing, not a league table · tag an engineer and the command runs with that person as its subject, written as preparation for a one-on-one

Threads that pick up later · every conversation is stored under your organization, so next week's follow-up continues the same thread instead of starting a new one

Commands run from the MCP connector too: the same slash command in the Larridin composer and the same question typed into Claude reach the same data under the same permissions.

A closer look

Engineering answers with the evidence attached

Ask AI connects a question to your organization’s engineering analytics. The MCP connector is in Beta and must be enabled for your organization; available tools depend on configuration and role.

Documented MCP clients and setup
Client	Connection	Requirements
Claude Code	Remote HTTP server; command in the setup guide	Larridin account, OAuth sign-in, organization enablement
Claude Desktop and Cursor	Remote server connection; bridge configuration where needed	Remote MCP and OAuth support, or the documented bridge
VS Code	Remote MCP server configuration	Remote HTTP and OAuth support; organization enablement

Illustrative question and answer

What did we get for $28,400 of AI spend?

In this illustrative four-week period, the team delivered 1,920 Engineering Output points: 67.6 points per $1K of AI spend. A separate estimate values engineering time saved at $164,400, giving $136,000 in estimated net value and a 4.8× estimated net ROI. Review quality and reliability alongside these figures.

Example inputs, not customer results. Engineering Output measures scored work; estimated time saved requires its own assumptions and cannot be derived from the Output score alone.

Review the calculation and assumptions

Questions about Ask AI

What data can Ask AI access?

Enabled analytics tools cover engineering velocity, quality, AI coding adoption and cost, agent effectiveness, and agent readiness. Available domains depend on the organization’s configuration and the caller’s role.

How are permissions enforced?

OAuth identifies the Larridin user and organization. Tool availability and analytics access are role-scoped. Confirm the enabled tools and their access scope with your administrator before rollout.

Can Ask AI change data or only answer questions?

The documented analytics tools query data rather than editing source systems. Asking a question does create or continue a saved Ask conversation under your organization’s retention terms.

How should I interpret the product examples?

The product views on this page use illustrative data to explain the metrics and workflows. For metric definitions, sample calculations, and assumptions, see our measurement methodology.

Ask your first question today.

Connect your engineering data, or enable the MCP connector to explore it from a supported assistant.

Get Pricing Try Now

Ask in plain English. Get the evidence back.