AI coding agents and agent orchestration

Jackhamr — Lets users spin up teams of AI agents for departmental work and pays only for compute and LLM usage.

TokenFish — Shows remaining Codex and Claude Code quota and reset times in a small animated desktop widget.

CrickCoder — Provides pay-per-use AI autocomplete and chat inside VS Code and JetBrains sidebars, billed at raw API cost plus a margin.

Médula — Checks each agent's file write against other agents' work in a shared repo, using a fast model to flag clashes before the write.

Amendary — Monitors GitHub code changes and drafts corrections to selected Notion or GitHub docs that have become outdated, showing the source evidence.

WhyVault — Hosts a shared Markdown knowledge vault that AI agents like Claude, ChatGPT and Cursor can read from and write decisions back to via MCP.

Between — Coordinates two AI coding agents through repository diffs, having a reviewer agent write structured review files that the developer agent reacts to.

Verb — Lets SaaS users type natural-language requests that an embedded AI agent carries out using their own permissions, with confirmation for data changes and an audit log.

Superbrainstorming — Provides a fork of the Superpowers agent workflow that keeps the brainstorming stage and skips the planning and test-driven stages, going straight from brainstorm to implementation.

Auth Your Agent — Lets an AI agent act on websites without ever seeing the user's password, requiring phone approval for logins and any data-changing actions.

GhosttyEXTREME — Shows the live status of multiple coding agent sessions in a terminal, with a file editor that follows agent edits, a repo map, and infrastructure views.

Freekit — Connects VS Code to existing web AI chat accounts through browser automation, or to a local Ollama model, to generate and apply code changes.

SCC Runner — Verifies an AI agent's claimed outcomes against real system evidence and returns a verdict.

Vibe Multiagent — Orchestrates multiple AI coding agents with role separation, safety hooks that block risky commands, and an offline HTML dashboard for monitoring their tasks.

Token Harness — Installs and connects token-reducing optimizers to coding agents and routes suitable subtasks to cheaper models, shown in a local dashboard.

Hermes Skill Deck — Browses, searches, and organizes AI agent skill files across profiles, with favorites and quick access to files in Finder.

devterm — Opens a real shell terminal panel inside the browser on any page and can paste shell code blocks into it for the user to run.

I let an AI agent take only the… — Takes small and medium tickets from a live SaaS backlog and produces code changes as pull requests for human review.

Talyn — Watches pull requests, sends an agent back to fix failed CI or merge conflicts, and merges PRs once they are green.

Orbtile — Shows local coding sessions on a Logitech keypad and a Mac menu-bar popover so users can see which sessions need attention and read their replies.

Savepoint — Guides AI-assisted coding through a local terminal workflow of planning, small tasks, and independent checks, storing everything as Markdown and YAML files in the repo.

Sloppy — Scans PHP code for common quality problems like silent exception catches, oversized methods, and queries inside loops, and can run during AI agent edits to report only newly introduced issues.

OMK — Runs a terminal coding agent that works on a local project, with multi-model switching and optional goal mode that checks a user-approved acceptance command.

Eigenwise Toolshed — Provides a collection of installable Claude Code plugins for repo maps, scoped rules, a local task board, model access, usage monitoring and workspace setup.

codeTAC — Traces a UI action back to the components, server functions, database calls and external services behind it.