The Tap Registry Grew Beyond Code: 108 Agents, 27 Domains, and Workflows That Verify Their Own Work

This post originally appeared on octomind.run/blog/octomind-tap-registry-expansion. Cross-posted to dev.to.

---

The cloud launch got the attention this week. But something quieter happened alongside it — the tap registry went through its biggest expansion since taps were first introduced, and it's worth a closer look if you're building with agents.

The default registry now ships 108 agents across 27 domains, 71 capabilities, and 18 workflows. All available with one command. No downloads, no marketplace accounts, no setup.

Here's what changed and why it matters.

What a tap actually is

If you haven't used taps before, the mental model is simple: a tap is a git-hosted registry of TOML manifests. Each manifest defines one agent — its system prompt, model, tools, MCP servers. Octomind fetches the manifest at runtime and merges it into your config.

Nothing gets installed. There's no global state to rot.

octomind run developer:rust      # a Rust specialist
octomind run lawyer:immigration  # an immigration-law explainer
octomind workflow harden         # a multi-step security audit pipeline

The tag format is domain:spec. The official muvon/octomind-tap ships as the default registry. You can browse it on the site, and adding your own registry is a single octomind tap away.

The point: agents should be distributed like code — reviewable in a diff, not installed like apps.

The registry isn't just for developers anymore

This is the biggest shift. The tap registry started as a developer toolbox, but most problems people bring to an agent aren't code problems.

This week added 17 new agents across domains you wouldn't expect in a dev tool:

• Family — family:parenting, family:eldercare, family:bereavement (an agent for the paperwork and decisions that arrive at the worst possible time, written with the gravity that moment deserves)
• Finance — finance:tax, finance:retirement, finance:insurance
• Health & wellbeing — doctor:sleep, coach:wellbeing, coach:negotiation
• Life — lawyer:immigration, home:realty, vet:general, vet:behavior, vehicle:mechanic, tutor:coding, content:fiction

These aren't replacements for licensed professionals. They're structured, well-prompted assistants that help you prepare, organize, and understand. The value is that the preparation — which is honestly 80% of the work — no longer starts from a blank page.

There's also assistant:concierge, which routes across the whole ecosystem. A bare octomind run discovers matching specialists by intent.

Four new workflows that close the loop

Workflows are multi-step TOML pipelines. Each step gets its own agent, model, and toolset, with validate scripts acting as deterministic quality gates between steps.

The shared pattern across all four new workflows: don't stop at the first answer — loop until the check passes.

harden

An OWASP-lens security audit of your current repo, then a fix ⇄ re-audit loop that runs until findings are resolved and your project's own checks stay green.

octomind workflow harden

upgrade

Dependency bumps — named packages, or safe patch/minor by default — running in a bump ⇄ verify loop until your build is green.

scout

This one's interesting. It's the phase before anyone has an idea: take a field or market, mine it for real evidenced pain through three blind parallel sweeps, and come back with vetted opportunities instead of vibes.

learn

Bootstraps a project knowledge base. Reads the codebase, extracts durable facts, promotes them into .box/ where they're indexed and readable by every agent that works on the project afterward.

All four run the same way as everything else — octomind workflow — reading from stdin, writing to stdout, scriptable by construction. That brings the registry to 18 workflows total.

Agents that remember the project

A new octobrain-backed knowledge capability gives agents durable project memory. assistant:knowledge curates it, the learn workflow seeds it from the codebase.

Facts get extracted once, stored in the repo's .box/ directory, and become available to every specialist that touches the project afterward. You run learn once. Every agent after that starts with context instead of zero.

The unglamorous fixes that actually matter

Two things that don't make headlines but fix real pain:

Dependency installs now work in containers and CI. Tap agents bootstrap their own tooling (Node for MCP servers), and those scripts assumed sudo exists. Root containers — Docker, CI images — don't have it. Every dep script now escalates through a helper that runs bare as root, uses sudo when present, and tells you exactly what's missing otherwise.

Web search needs no API key. The websearch capability defaults to DuckDuckGo now. One less credential between a fresh install and a working researcher agent.

Use it anywhere

Everything works on a local Octomind install (one curl command) and comes preloaded on cloud machines. Models route through the hub with octohub:auto routing.

What to do next

• Browse the full registry at octomind.run/tap — see what's available before you build something from scratch
• Read the manifests at github.com/Muvon/octomind-tap — they're TOML files, fully reviewable, and contributions are open
• Try a workflow — octomind workflow harden on a repo you care about and see the loop-until-green pattern in action

If there's a specialist you keep wishing existed, it's one manifest away. And if you build one, we'd genuinely like to see it.