AI Agents

turri

A coding agent that rewrites code from specs instead of patching it

It puts a real tool behind the idea that specs and tests should outlive code, so you can see whether regeneration holds up on your project.

cogit

Records each prompt to your coding agent so any line of code can explain itself

When an agent wrote the code, the reasoning lives in a chat you closed; cogit keeps it beside the commit so the next person can ask why.

Taster

A guard layer that checks agent tool calls and screens what tools send back

If your agent touches customer systems, this puts your rules outside the prompt, where a long context or a poisoned web page cannot talk it out of them.

Hubzoid

Self-hosted team agents whose knowledge and rules live in plain files

It lets a small team share one governed agent brain with sign-in and approvals, using Claude or Codex subscriptions they already pay for.

docxtor

Lets AI agents edit Word files with tracked changes and nothing else touched

If your agent edits contracts or reports, this keeps every untouched byte of the .docx intact and records changes as tracked revisions.

klad

A Rust painting and animation app where you and an AI agent share one canvas

Because every stroke stays editable data, an agent's drawing can be corrected line by line instead of regenerated from scratch.

Modern Web Guidance

Browser-vendor guides that teach coding agents today's built-in web APIs

If agents write your front end, this cuts the hand-rolled modals, polyfills and div soup they reach for by default, with support tables so you can weigh each feature against your users.

Willder

Shared, permissioned memory for coding agents, with an audit row per read

When several agents and people work from the same knowledge, Willder gives each one a scoped, revocable token so you can see and cut off exactly what any agent touches.

Thrum

A local message bus that lets your coding agents review each other's work

If you already pay for several coding agents, Thrum lets them coordinate on your own machines, so you answer one agent instead of supervising five.

STE80 Humanizer

An agent skill that edits LLM prose toward Simplified Technical English

If your docs or changelogs are drafted by a model, this borrows an aerospace writing standard to make them plainer, and costs one install command to try.

Anyone using both Claude Code and Codex stops editing two config files by hand and stops leaving API tokens in plain text.

agent-email-guard

Rate limits, duplicate checks and a kill switch for email sent by AI agents

If an agent in your product can send email, this is a cheap seatbelt against it spamming a customer when a timed-out call gets retried with new wording.

glassbrain

A browser viewer and diff tool for the knowledge graphs your agents write

When an agent keeps its memory and plans as a plain JSON graph, you can review what changed between runs the way you review code.

rea-video

An agent skill that turns a reference video into a remake plan

When a client sends a clip and asks for something similar, rea-video gives your agent a shot list that separates real footage from AI and motion graphics, so you can scope it.

Craft Anything

A Rust engine that studies an app, then has a coding agent rebuild it

It turns 'how did they build that?' into a working clean-room prototype, so you can study a competitor's mechanics without copying their code.

Chronosphere

Replays a pull request as the ordered steps that could have built it

When an agent hands you an enormous pull request, reading it as a sequence of small, explained steps is a more honest review than skimming one giant diff.

Permaura

A credential vault and approval gate between AI assistants and your apps

If you let Claude Code or Cursor touch Stripe or GitHub, it keeps the keys out of the agent and puts risky actions behind an approval you sign on a device.

BLA

Hardens your shipped builds so AI reverse-engineering agents waste effort

If you ship an Electron app, a JS bundle or a binary, AI agents can now pick it apart cheaply, and BLA is a free way to make that cost real again.

Imza

A tiny native terminal agent where you co-write the spec before any code

It suits developers who want an agent to agree a written plan first and to touch files only through one permission-checked Lua sandbox.

tstack

Nine agent skills that carry a pull request from first commit to merge

If your agents already write the code but you still shepherd every PR through CI, review bots and merge, these skills show one heavy user's way of handing that off.

LDraw Mecha

Turns an LDraw Lego model into a rigged, moving Blender scene using an AI agent

It shows agents doing spatial reasoning on a real file format: working out which parts of a model move, how they connect, and writing the Blender code to prove it.

rea-video

An agent skill that works out how a video was made and how to remake it

If you study competitors' launch videos or ads, this turns a guess about how they were made into a shot list you can actually brief.

flawless-ui-ux

Gives your coding agent design judgement, then checks the pages it builds

Agent-built interfaces tend to look alike and miss basics; this gives the agent a brief to follow and a check to run before you review the result.

peekabot

One board for your Claude Code sessions in tmux: who needs you, who is done

If you run a dozen or more agent sessions, it tells you which one is blocked on you right now, without any hooks or a background daemon.