We're on Product Hunt: Why We Built an AI Teammate Instead of Another Dashboard

Tandemu is live on Product Hunt. An open-source AI teammate with persistent memory, automatic DORA metrics, and commit-level AI attribution — built for developers, not against them.

Tandemu is live on Product Hunt. Before you scroll to the upvote button, here’s what it is and why it exists.

The Problem We Kept Running Into

Engineering managers live in a gap. On one side, you have developer productivity dashboards — Waydev, Jellyfish, LinearB — that track IDE telemetry, count commits, and generate scores your developers resent. On the other side, you have AI context tools that help developers work faster but give management zero visibility into what’s actually happening.

You end up choosing between data your team doesn’t trust and tools your leads can’t use.

We built Tandemu because we didn’t want to choose. The question was straightforward: can you give developers a better AI coding experience and give leads real metrics from the same system, without any of it feeling like surveillance?

What Tandemu Actually Does

Tandemu runs alongside Claude Code. Developers use terminal skills — /morning to pick a task, /finish to wrap up — and everything between those two points gets measured automatically. No timesheets. No surveys. No browser extensions.

Here’s what’s under the hood.

Persistent AI Memory

Every AI coding session generates context — architectural decisions, debugging patterns, personal coding preferences. Most tools throw that away between sessions. Tandemu keeps it. Your AI remembers your codebase, your style, and your team’s conventions. It gets better the more you use it.

The memory is scoped: personal preferences stay personal, architectural knowledge gets shared across the team. Leads can browse the memory dashboard to see knowledge gaps — which modules have coverage and which ones the AI is flying blind on.

Commit-Level AI Attribution

When Claude Code edits a file, Tandemu records which tool fired, which file it touched, and how many lines it wrote. That’s not inference from IDE telemetry. That’s native OpenTelemetry instrumentation of the AI’s actual tool calls.

The result is a per-file, per-developer AI ratio — the percentage of code the AI wrote versus what the developer wrote manually. You can segment it by directory, by task, or by time window. A 75% AI ratio on test boilerplate is healthy. A 75% AI ratio on your payment processing logic is a conversation worth having.

Automatic DORA Metrics

Deployment frequency and lead time are calculated from task completions — no CI/CD integration required on day one. Every /morning to /finish cycle is a measured unit of delivery with known duration, known code changes, and known AI involvement.

Friction Detection

When the AI gets stuck — prompt loops, tool errors, repeated failed attempts on the same file — Tandemu flags it. Friction is classified by severity (high, medium, low) and surfaced on a heatmap. The modules with the most friction are usually the ones that need better documentation or architectural context, not more developer effort.

Zero-Ceremony Workflow

Standups, timesheets, and status updates are generated from telemetry data. The /standup skill produces a team report from what actually happened — tasks completed, time spent, friction encountered — without anyone writing a summary or sitting through a meeting.

Open Source, Free for Solo Developers

Tandemu is open source. You can self-host the entire stack — NestJS backend, PostgreSQL, ClickHouse for telemetry, Next.js dashboard — with a single Docker Compose file. The self-hosted version has full feature parity with the cloud version. Your data never leaves your infrastructure.

For individual developers, the cloud version is free. You get persistent AI memory, the terminal workflow skills, and your own metrics dashboard at no cost.

For teams and organizations, the SaaS plan adds team dashboards, cross-developer memory, friction heatmaps, DORA metrics, and integrations with Jira, Linear, ClickUp, GitHub Issues, and Asana.

What Makes This Different

Most productivity tools are built for managers and tolerated by developers. Tandemu is the opposite. The primary user is the developer — the AI teammate, the memory, the workflow skills. The management layer is a byproduct of the developer using the tool, not the other way around.

That inversion matters. When the tool makes the developer’s day better, adoption isn’t a problem. When adoption isn’t a problem, the data is real. When the data is real, leads can actually trust it.

Try It

We’re live on Product Hunt today. If you manage an engineering team that uses AI coding tools and you’re tired of choosing between developer surveillance and management blindness, check it out.

If you want to self-host first, the Docker setup guide takes about ten minutes. If you want to jump straight in, create a free account and connect Claude Code.