
Proletariat Cli
io.github.chrismcdermutPublished by GitHub account chrismcdermutv0.3.134Updated Oct 28, 2026
Agent orchestration platform - multi-agent spawning, tickets, boards, and workflows
Overview
Local MCP server for prlt, an agent orchestration CLI that lets an assistant create tickets, spawn coding agents and track their work.
- What you can do
- Create, list, view, edit, move, link and delete tickets, epics, projects and specs
- Spawn, batch-spawn, watch, complete or revise agent work on tickets, and groom requirements
- Manage agents, naming themes, executions, tmux sessions and Docker containers
- Create branches and pull requests, link PRs to tickets, and check GitHub auth status
- Read and watch kanban boards, workflows, phases and statuses
- Run reusable action templates and workspace, repo and utility commands
- When to use it
- Useful when you want an assistant to drive a multi-agent development workflow from one place: filing tickets, launching isolated coding agents on branches, and following their pull requests through review. It suits teams already working in git repositories who want structured, ticket-based context instead of freeform chat.
- Requirements
- Runs as a local process, typically launched as prlt mcp-server or via npx with the @proletariat/cli package. Needs Node.js 20 or newer, Git, SQLite and tmux; Docker is optional for isolated execution. Claude Code must be authenticated separately, and GitHub operations use a GITHUB_TOKEN environment variable.
Installation
In SourceWeft
- Open Proletariat Cli in the dashboard and add it to a workspace.
- Enable the server for the chats that should use its tools.
Desktop only via STDIO. STDIO servers start a local process, so they need the SourceWeft desktop host.
Other MCP clients
Follow the launch instructions in the repository.
README
Seize the means of production - Ship 100x.
[npm] [MCP Registry] [License]
Agent orchestration platform for AI labor. Spin up workers for all work, on demand.
Themed agents, including billionaires - Finally, they work for us.
⚠️ Beta Software — Under active development. Commands and APIs may change between versions, and bugs are actively being squashed.
Let's get you shipping Book a call - I'm happy to help you get prlt running or chat feedback, ideas, multi-agent workflows, and the future of work/labor (and economic labor theory..)
TLDR
prlt is an agent orchestration platform for AI labor. Spin up workers on demand, coordinate multi-agent development from one CLI. Isolated workspaces, secure containers, persistent state.
Agent spawns in its own branch, writes code, opens PR. You review and merge.
Why prlt?
- Isolated - Each agent gets its own git branch. No conflicts.
- Secure - Docker containers, sandboxed from your host.
- Durable - Tmux sessions persist. Close window, agent keeps working.
- Trackable - One database, one CLI, all your agents.
- Ephemeral - Spawn agents on demand. They work, they PR, they're done.
- Structured - Tickets provide structured context, not freeform chat.
- Persistent - Tickets accumulate context over time. Hand off between agents.
- Agent-native -
--jsonmode lets AI agents drive the CLI programmatically.
Quick Start
Workflow Diagram (click to expand)
Spawn agents to implement, groom, or review—not just write code.
Interactive Menus
prlt work guides you through project and ticket selection:
Choose your operation—start a single agent, batch spawn, or watch a column:
Select tickets to spawn, grouped by priority:
Deep Dive
Installation
Homebrew (recommended)
Works on both Apple Silicon (arm64) and Intel (x86_64) Macs. No compiler needed.
Upgrade:
npm / pnpm (all platforms)
pnpm 10+ note: pnpm 10 blocks native addon build scripts by default. If
prltfails with a native module error after install, runpnpm approve-buildsin the global store and reinstall, or usenpm/brewinstead.
Verify:
MCP Server
prlt includes a built-in MCP server with 100+ tools. Add it to your AI client:
Claude Code (~/.claude.json):
Cursor / Other clients (via npx):
Listed on: MCP Registry | npm
Data Model
Example Workflow
A workflow defines how tickets move through your process. Projects reference a workflow, and multiple projects can share the same one.
Tickets flow through statuses as work progresses. Agents automatically move tickets when they start work, open PRs, or complete tasks.
Workspace Structure
Each agent gets a copy of all repos (repo scoping coming soon). Work happens on isolated branches.
Agent Naming Themes
Themes control how agents are named. Staff agents use theme names directly (e.g., bezos, camry). Ephemeral agents add an adjective prefix (e.g., bold-bezos, keen-camry). Currently ephemeral names also include a number suffix (bold-bezos-1), but this will be removed soon.
Built-in Themes:
billionaires — Finally, they work for us.
Theme Commands:
Themes are selected during prlt new.
Three Ways to Use Commands
1. Interactive (Humans)
Run without flags—get guided prompts:
View ticket details with prlt ticket:
2. JSON Mode (AI Agents)
Add --json for machine-readable output:
AI agents parse this, make selections, call the next command.
3. Flags (Scripts/CI)
Pass everything directly:
Execution Modes
Environment - where the agent runs:
Display - how you see it:
Permissions - agent access level:
All sessions run in tmux under the hood—close the window, agent keeps working.
Parallel Agents
Work on multiple tickets simultaneously.
Interactive (humans):
JSON mode (AI agents): (multi-select WIP)
Flags (scripts/CI):
Each agent works in its own branch. No conflicts.
Scaling: The main limit is your machine. 50+ concurrent agents is achievable—depends on CPU, RAM, and whether you're running Docker or host mode.
Monitor running agents with prlt execution:
Agent-created PRs ready for review:
Command Reference
Full Command Reference (click to expand)
Run prlt <command> --help for flags and options.
Use Cases
Parallel Feature Development
Three agents, three branches, three PRs. You review and merge.
Bug Bash
Grooming Session
Have an agent refine ticket requirements:
Agent adds acceptance criteria, subtasks, estimates.
Environment Variables
Claude Code handles its own authentication via claude login.
Requirements
- Node.js 20+ (22 LTS recommended)
- Git
- Claude Code (
claude loginto authenticate) - SQLite
- Tmux (session persistence)
- Docker (optional—for isolated execution)
Troubleshooting Installation
bun install fails on better-sqlite3
Symptom: bun install -g @proletariat/cli fails with isexe or node-gyp errors during the better-sqlite3 native module build.
Cause: Bun's node-gyp compatibility is limited. The which dependency inside node-gyp uses isexe, which is incompatible with Bun's runtime shims.
Fix: Use npm or Homebrew instead:
If you must use Bun, ensure Node.js 22 (LTS) is also installed and set better-sqlite3 to use its prebuilt binaries:
npm EACCES: permission denied
Symptom: npm install -g @proletariat/cli fails with EACCES: permission denied on /opt/homebrew/lib/node_modules or /usr/local/lib/node_modules.
Fix: Configure npm to use a user-writable directory:
Or use Homebrew instead (macOS):
Native module errors after install
Symptom: prlt runs but crashes with better_sqlite3.node or ABI mismatch errors.
Fix:
See the full troubleshooting guide for more details.
Support
- Discord: discord.gg/tmZyjNNSvw
- GitHub Issues: Report bugs or request features
- Setup Help: Book a call - I'll help you get things running
License
Apache 2.0
[npm version] [Downloads] [License]
Star on GitHub | Install from NPM | Report Issues
Made with ⚒️ by the proletariat.
Source: apps/cli/README.md at commit 0f0f9fa
Tools
0Version history
1- v0.3.134LatestSep 16, 2026
Score
50% of the applicable checks met
How complete this listing is, computed from the registry entry, the endpoint and public sources. Not a rating of the server and not a statement that it is safe or working today: reachability is a dated check, and no installs or reviews go into the score.
- Reached in the last 30 days25 pts
Keep the endpoint answering the MCP handshake; SourceWeft probes it every 30 days.
- Auth declared10 pts
Declare the credential headers in the registry entry, or keep the endpoint answering the probe so its auth can be classified.
- Tools listed15 pts
Answer tools/list without a credential, or ship a package that starts in a sandbox without one.
- DescriptionRequired15 pts
- Logo10 pts
- Documentation15 pts
- License10 pts