Quota

Overview

At Tandem we back early-stage companies, and part of that work is helping them actually sell. Not just advising on go-to-market, but running pipeline: understanding the buyer, writing the outreach, working the sequence. The problem is that doing that well is intensely personal, and personal doesn’t scale the way a spreadsheet does.

Quota is my answer to that. It’s a team of AI agents, built on Claude, that handles everything about sales that doesn’t require human judgment, the research, the enrichment, the first drafts, and hands the rest back to you. I designed and built it, we run it at Tandem, and I open-sourced the framework.

The problem

Every team that tries to scale outreach lands in one of two places. Either a firehose of templated garbage that burns the list fast, or a team that knows the generic approach is wrong but never has time to do the research to do it right. The best deals I’ve seen close came from someone showing up with real context: knowing the prospect’s business, their likely pain, their competitive landscape, and opening a conversation that felt written for them.

I wanted a third option. What if the research, enrichment, and first-draft work just happened overnight, so that every morning there was a short list of the accounts most worth my time, each with a brief and a draft already waiting to review?

What I built

Quota runs a team of eight agents on a daily schedule, orchestrated by a CRO agent that reads the current OKRs each morning and decides what the day should focus on.

• CRO reviews the OKRs at 7am, dispatches the other agents, and posts a plan to Slack.
• Scout researches companies and finds and verifies contacts.
• Outreach executes three-touch email sequences. Tier-1 opportunities land in Gmail drafts for you to review, edit, and send. Lower tiers move faster on their own.
• Enablement writes call-prep briefs from CRM data before meetings.
• Channels reports on partner and referral opportunities.
• Inbox reads replies every fifteen minutes, classifies them, and updates the CRM.
• Digest posts a pipeline summary each morning.
• Follow-Up drafts responses and queues the important ones for approval.

Everything surfaces through a dashboard and a Slack integration, so approvals come to you instead of asking you to open one more app. The agent never sends a Tier-1 email without you in the loop.

How it works

Claude does the reasoning. A Postgres database holds the agent configs, OKRs, and run history. Attio is the CRM and the shared state that every agent reads from and writes to, so nothing lives in a silo. Gmail handles send, draft, and inbox. Apollo and FullEnrich handle contact discovery and verification. Slack handles approvals. The whole thing is built to run in Claude Code and deploys in a couple of hours.

Two decisions shaped the build more than any other.

First, every agent’s prompt is a plain Markdown file, hot-reloadable from the dashboard with no redeploy. The prompt layer is where all the judgment lives, what good outreach looks like, what to research, how to write a brief, so it had to be the easiest thing in the system to change, not something buried in code.

Second, everything degrades gracefully. Skip an optional integration and the server still starts; the agents that need it skip those steps and log a warning instead of falling over. A system that runs unattended every morning can’t be brittle.

Open-source, with a line drawn

I open-sourced the whole framework: the agent architecture, the React dashboard, the Gmail and Slack integrations, the approval flow. What I kept private were the specific prompts we run at Tandem. Those are the real IP. The prompts in the repo are instructional templates that teach you how to write your own, with a copy-paste Claude prompt that generates a business-specific version from your context.

This is not a SaaS you rent. It’s a system you own and run, and the layer that holds all the value is the one you control.

What it costs

Running it is roughly $90 to $180 a month, all in: Claude API (Sonnet for the nuanced work like drafting and research, Haiku for the cheap jobs like inbox triage), a CRM, contact tooling, and about $10 for hosting and Postgres on Railway. Exact token spend per run shows up in the dashboard.

What surprised me

I expected it to save time. It did. What I didn’t expect was that it made the outreach better. When a fully researched brief is sitting in front of you before you write, you don’t start from a blank page, you start from a better one. The agents do the legwork and you do the craft. The relationships we care about get more attention now, not less. That was the point.

Links

Quota is open-source on GitHub at github.com/darrellwhitelaw/quota. I also wrote about the thinking behind it in Meet Quota.