AI Launchpad

A lightweight framework for making simultaneous agentic updates to multiple repositories.

Initial Setup

The first time you are setting up this project, you will need to:

1. Install uv
2. Run uv sync to create a virtual environment and install all dependencies
3. Copy config-example.toml to config.toml and fill in your configuration and credentials

Optional config:

• expected_source_repo_branch (example: "dev"): if set, Launchpad will pause before creating each worktree until the source repo is on this branch.

Quickstart

Once this project is set up, you can get started with:

uv run python launch.py -h

Sources

Local TODO File Sources

You can load work items from a local text file. Each line that starts with - is treated as a work item title; all other lines are ignored. The file's parent directory is used as the relevant source directory for the items it contains.

# Load work items from a local todo file
uv run python launch.py --todo-file ./todo

# The flag can be repeated to combine multiple files
uv run python launch.py --todo-file ./todo --todo-file ../other-repo/todo

Example todo file:

- Add retry logic to the upload client
- Fix the off-by-one error in the pagination helper

GitHub Issue Sources

You can load work items directly from GitHub issues:

Set github.access_token in config.toml if queries need private-repo access.

# Pull issues using a GitHub issue query for one repo
uv run python launch.py --github-issue-query "repo:owner/repo is:open label:bug sort:updated-desc"

# Launch a single issue by number
uv run python launch.py --github-issue-query "repo:owner/repo is:issue 123"

# Mix and match multiple sources
uv run python launch.py \
  --todo-file ./todo \
  --github-issue-query "repo:owner/repo-a is:open assignee:@me" \
  --github-issue-query "org:owner is:open label:bug"

The query uses GitHub's issue search syntax. To target a single issue, include the issue number directly in the query string along with the repo qualifier.

Jira Sources

You can load work items from Jira using JQL:

uv run python launch.py --jira-jql "project = CORE AND status = 'Backlog' ORDER BY created DESC"

# Launch a single Jira ticket by key
uv run python launch.py --jira-jql "key = CORE-123"

Required config for Jira (set in config.toml):

• jira.org_name
• jira.email
• jira.api_token

Linear Sources

You can load work items from Linear:

# Launch a single Linear issue by identifier
uv run python launch.py --linear-query "ENG-123"

# Pull issues using a Linear IssueFilter as JSON
uv run python launch.py --linear-query '{"team":{"key":{"eq":"ENG"}},"state":{"type":{"neq":"completed"}}}'

The query is either an issue identifier (e.g. ENG-123) or a Linear IssueFilter encoded as JSON.

Required config for Linear (set in config.toml):

• linear.api_key — a Linear personal API key