Open Source · MIT Licensed

Claude Code
becomes a task engine.

Drop a YAML file. TaskSmith assembles your context, invokes Claude Code, validates the output, retries until tests pass, and pings you when it's done.

$ npm install -g tasksmith-cli [CLICK TO COPY]
then: tasksmith setuptasksmith run
~/my-api — tasksmith v0.3.1
tasksmith submit -t ralph-loop \
-p "Add input validation to /users" \
--project my-api
 
■ TASK SUBMITTED
ID: task-20250217-a3f9c1
Template: ralph-loop
Model: sonnet
Queue: 1 of 1
 
// 3 minutes later, your phone buzzes:
✓ COMPLETE — passed after 2 iteration(s)
npm test ■ ALL PASSING
sys.capabilities

Everything Claude Code
should have been.

The orchestration layer that turns a single-shot CLI into an autonomous coding agent.

01 // CONTEXT

Compiled Prompts

SOUL.md, USER.md, conventions, memory, and project context automatically assembled into every prompt. Claude always has full context. You never repeat yourself.

02 // PARALLEL

Worker Pool

Run multiple tasks simultaneously. Configurable concurrency with priority queuing. Each task gets its own git worktree — auto-opens a PR on success, discards on failure.

03 // VALIDATION

Ralph Loop

Define a validation command. TaskSmith runs Claude Code, checks output, feeds errors back, retries. Iterate until npm test passes.

04 // PERSISTENCE

Three-Tier Memory

Hot memory loads every prompt. Warm JSONL logs are searchable. Cold archives compressed. Semantic search via local or cloud embeddings. Claude remembers across tasks.

05 // COMMS

Notifications Everywhere

Discord, Slack, ntfy.sh, email, webhooks. Know when tasks finish. Get results on your phone while doing something else.

06 // SCHEDULE

Cron & Plugins

Scheduled tasks via cron. 8 official plugins ship bundled — GitHub, Metrics, Docker, JIRA, Postgres, Proxmox, Cloudflare, Semantic Memory. Or build your own.

task.pipeline

How a task flows.

From YAML to passing tests in minutes, not hours.

01

Submit

YAML file, Discord message, REST API, or CLI command

02

Assemble

Soul + user + conventions + memory + project into one prompt

03

Execute

Claude Code runs with full context in your project directory

04

Validate

Run test suite. If it fails, feed errors back and retry

05

Notify

Push to Discord, Slack, phone. Archive to memory

<5k
Lines of core TypeScript
0
External frameworks
8
Official plugins
10
Notification providers
plugins.official

8 plugins.
Zero installs.

Official plugins ship with the CLI. Enable any with one line in config. Lazy-loaded — disabled plugins cost nothing.

# tasksmith.yaml
plugins:
  - github           # Issues/PRs on fail/success
  - metrics          # Execution analytics
  - docker           # Container isolation
  - jira             # Ticket integration
  - postgres         # SQL task history
  - proxmox          # VM provisioning
  - cloudflare       # Pages deploy/rollback
  - semantic-memory  # Vector search

GitHub

Auto-create issues on failure. Comment results on linked issues. Close on success. Just set GITHUB_TOKEN.

📊

Metrics

Success rates, timing, model/template breakdowns. tasksmith metrics for a dashboard in your terminal.

🔍

Semantic Memory

Vector-based search over task history via Ollama (local), OpenAI, or Gemini embeddings. tasksmith semantic --query "auth refactor"

Cloudflare

Deploy to Cloudflare Pages on task success. Rollback, cache purge, deployment history. tasksmith cf --action deploy

+

Build Your Own

tasksmith plugin create my-thing
npm is the plugin manager. Publish to @tasksmith-dev/* or tasksmith-plugin-*.

SysOp

Matt

Software architect from Birmingham, AL. Built TaskSmith because OpenClaw's 430k lines were too many and Claude Code's zero orchestration wasn't enough. Under 5,000 lines of core TypeScript — every module fits in your head. 8 bundled plugins and a full plugin system for everything else.