Ultrametric open-sources a repository of startup workflows

Running a startup with AI agents usually means a dozen tabs: a cap table in one, a vendor comparison in another, a hiring doc somewhere else. Ultrametric gathers that material into one public repository built for founders who use agents to get real work done, not just to ship code.

The code is open source under Apache-2.0. The ranking datasets carry separate reuse restrictions, so check those before you republish any of the vendor comparisons.

What's in the Repository

Ultrametric is a public collection of startup workflows, vendor rankings, calculations, and documents. It connects each business process to the vendors, math, and paperwork that step actually needs. The repository is organized around five areas:

• Step-by-step processes that show whether a task needs an agent, a form, or a human, with timing and reversibility noted for each step.
• Agent-access rankings that compare banking, payroll, CRM, and other vendors by how well an agent can access and control them.
• Runnable startup math covering cap tables, SAFE conversion, dilution, runway, and vesting, with source-cited formulas and worked-example tests so you can check the numbers yourself.
• A document index of published formation, fundraising, hiring, and commercial documents, grouped by use case and listed with source links and last-checked dates.
• A founder reading library of essays, guides, books, and talks sorted by stage, from deciding to start through acquisition.

Why It Matters

The useful part is the honesty about what agents cannot do. Knowing which step needs a human, and which vendors an agent can actually reach into, saves you from automating the wrong thing first. The math and the document index come with their sources and check dates attached, so you can see how current an answer is instead of trusting a stale template. If you are building with agents and want the boring operational side laid out in one place, this is worth a look.