Agent Atlas

Practice

The Tool Shelf

The agents are the same for everyone. The tools around them are where builders get clever. Thirty small open-source tools, each solving one problem you will hit once you run more than one agent.

One maker, fourteen tools

Jeffrey Emanuel builds the whole kit around his agents. Three of his tools run on this site's own machine.

One maker, fourteen tools

Jeffrey Emanuel

On GitHub as Dicklesworthstone. He builds the whole kit around his agents: a guard against destructive commands, search over old sessions, a mailbox between agents, a task graph viewer, a tmux controller. His own answer for coordinating many agents is four of them: Agent Mail, beads_rust, beads_viewer and ntm (Jeffrey Emanuel on X).

This site's own machine runs three of them: dcg blocks risky commands, cass searches old sessions, and cass memory turns them into notes.

destructive_command_guardruns here

A hook that blocks destructive shell commands before an agent runs them.

The Solution: dcg is a high-performance hook that intercepts destructive commands before they execute, blocking them with clear explanations and safer alternatives.

★ 6.4k · last commit 2026-11-14

mcp_agent_mail

A mailbox agents use to message each other and to reserve files.

A mail-like coordination layer for coding agents, exposed as an HTTP-only FastMCP server.

★ 2.2k · last commit 2026-11-01

beads_viewer

A terminal view of a Beads task graph that ranks what to do next.

The elegant, keyboard-driven terminal interface for the Beads issue tracker.

★ 1.7k · last commit 2026-11-14

coding_agent_session_searchruns here

Search across the coding-agent sessions already saved on this machine.

Unified, high-performance TUI to index and search your local coding agent history.

★ 1.2k · last commit 2026-11-14

beads_rust

Jeffrey Emanuel's Rust port of the older Beads task tracker.

A Rust port of Steve Yegge's beads, frozen at the "classic" SQLite + JSONL architecture I built my Agent Flywheel tooling around.

★ 1.2k · last commit 2026-11-13

cass_memory_systemruns here

Turn those saved sessions into memory the next agent can read.

Transforms scattered agent sessions into persistent, cross-agent memory—so every agent learns from every other agent's experience.

★ 465 · last commit 2026-11-06

ntm

A tmux controller that starts and watches several coding agents at once.

NTM turns tmux into a local control plane for multi-agent software development.

★ 464 · last commit 2026-11-14

ultimate_bug_scanner

A bug-pattern scanner an agent can run as a check.

One command to catch 1000+ bug patterns (always main, cache-busted)

★ 317 · last commit 2026-11-14

mcp_agent_mail_rust

The maintained Rust mailbox for the same agent coordination.

A mail-like coordination layer for AI coding agents, exposed as an MCP server with 45 tools and 25 resources, Git-backed archive, SQLite indexing, an interactive 16-screen TUI, a server-rendered web UI, and an agent-first robot CLI.

★ 253 · last commit 2026-11-14

meta_skill

A local store of reusable skills an agent can search.

Meta Skill (ms) is a local-first skill management platform that turns operational knowledge into structured, searchable, reusable artifacts.

★ 214 · last commit 2026-11-13

coding_agent_account_manager

Swap a coding agent's logged-in subscription when that plan hits its limit.

Sub-100ms account switching for AI coding CLIs with fixed-cost subscription plans. When you hit usage limits on Claude Max, GPT Pro, or Gemini Ultra, don't wait 60 seconds for browser OAuth—just swap to another account instantly.

★ 213 · last commit 2026-11-04

coding_agent_usage_tracker

One command that shows provider usage across coding-agent plans.

Coding Agent Usage Tracker — monitor LLM provider usage from a single CLI

★ 92 · last commit 2026-10-14

slb

A risky command waits for a peer's approval before it runs.

When an agent wants to run something risky (e.g., rm -rf, git push --force, kubectl delete, DROP TABLE), slb requires peer review and explicit approval before execution.

★ 82 · last commit 2026-11-07

post_compact_reminder

After Claude Code compacts the chat, a hook tells it to re-read AGENTS.md.

The Solution: A hook system that detects compaction events and injects a short plain-text reminder telling Claude to re-read AGENTS.md before doing anything else. One install, zero maintenance, works globally across all your projects.

★ 55 · last commit 2026-10-14

Memory between sessions

Turn old sessions into notes the next agent reads. 1 tool.

claude-mem

by Alex Newman

Remembers a coding session so the next one can use it.

Your coding agent forgets everything between sessions. Claude-Mem remembers for it.

★ 117k · last commit 2026-11-13

Agents working together

Several agents, one codebase, without stepping on each other. 3 tools.

gastown

by Steve Yegge

A workspace that coordinates several coding agents working on different tasks.

Gas Town is a workspace manager that lets you coordinate multiple AI coding agents (Claude Code, GitHub Copilot, Codex, Gemini, and others) working on different tasks.

★ 19k · last commit 2026-11-12

claude-squad

by smtg-ai

A terminal list of agents, each task in its own git workspace.

Each task gets its own isolated git workspace, so no conflicts

★ 8.8k · last commit 2026-11-13

agents-deck

by Grigorii Merkushev

A status board of the coding agents already running, including what each one costs.

Mission control for your Claude Code and Codex agents

★ 324 · last commit 2026-11-01

A task list agents share

One list of work on disk, instead of plans scattered in chats. 1 tool.

beads

by Steve Yegge

A task graph on disk that agents share as one list of work.

Beads provides a persistent, structured memory for coding agents. It replaces messy markdown plans with a dependency-aware graph, allowing agents to handle long-horizon tasks without losing context.

★ 29k · last commit 2026-11-13

Terminals built for many agents

Keep sessions alive, and get told when one needs you. 2 tools.

herdr

by herdrdev

A terminal runtime that keeps agent sessions running after you disconnect.

herdr keeps terminals running in a background server when you close the client or lose your SSH connection.

★ 51k · last commit 2026-10-29

cmux

by Austin Wang

A Mac terminal, built on Ghostty, with vertical tabs and notifications for coding agents.

A Ghostty-based macOS terminal with vertical tabs and notifications for AI coding agents

★ 29k · last commit 2026-11-13

One worktree per task

Each agent gets its own copy of the code. 3 tools.

worktrunk

by Maximilian Roos

A git-worktree command line made for running agents in parallel.

Worktrunk is a CLI for git worktree management, designed for running AI agents in parallel.

★ 13k · last commit 2026-11-13

workmux

by Raine Virta

Opens a tmux window and a git worktree together for each task.

No stashing, no branch switching, no conflicts.

★ 3.0k · last commit 2026-11-13

treehouse

by Kun Chen

Keeps a pool of worktrees so the next agent gets an environment that is already there.

Treehouse helps you manage a pool of reusable, isolated worktrees so each of your agents gets its own environment instantly — no cloning, no conflicts, no coordination overhead.

★ 1.9k · last commit 2026-11-11

Better instruction files

Keep AGENTS.md short and correct. 2 tools.

agents-md-cookbook

by Taiizor

Templates and a linter for a short AGENTS.md.

The tested, tool-agnostic AGENTS.md kit.

★ 29 · last commit 2026-08-08

agentlint

by Joseph O'Brien

A linter for agent harness files such as agents, skills, commands, hooks, and settings.

Linter for AI coding agent harness files.

★ 0 · last commit 2026-09-21

Know what you spend

Usage, quotas and the cost of each session. 1 tool.

ccusage

by ryoppippi

Reads local session logs and estimates what the tokens would have cost.

Analyze coding (agent) CLI token usage and costs from local data.

★ 19k · last commit 2026-11-14

And the rest

Loops, skills, checks and account switching. 1 tool.

ralph-tui

by Ben Williams

Runs an agent through a task list in a loop until the list is done.

AI Agent Loop Orchestrator - A terminal UI for orchestrating AI coding agents to work through task lists autonomously.

★ 2.5k · last commit 2026-11-04

How we picked

Factual
Active in the last six months, quotes from each README at a pinned commit, stars as one snapshot.

Grok searched GitHub on November 13, 2026 and skipped anything without a commit in six months. Every quote is from the tool's own README at a pinned commit, and Claude checked each one. Stars and dates are one snapshot; they move daily. Being listed is not an endorsement. Read the code before you give a tool your shell.