redline
llms.txt27 starsChanged 50 days agoApache-2.0View on GitHub
A local, budget-aware job dispatcher that runs queued coding-agent jobs in Claude Code, Codex CLI, Pi, Hermes, or a custom command when Codex or Claude subscription allowance is running behind pace or would otherwise expire. Redline watches 5-hour and weekly subscription allowances locally, explains each scheduling evaluation as RUN, WAIT, or fail-closed UNKNOWN, and dispatches eligible jobs, preferably into isolated Git worktrees or DevX sessions (Hermes runtimes and explicitly approved existing directories are also supported). It runs as a loopback…
The file
We split this llms.txt into the rules it gives an agent
10 rules. About 0 are common, 9 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈9
Distinctive · estimate
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 9
Read and follow the user's repository instructions first.
- Distinctive: line 11
Ask before replacing an existing configuration, and before installing or authenticating another harness or tool.
- Distinctive: line 12
Never request, print, or store account passwords or app-specific passwords.
- Distinctive: line 13
Never expose the loopback service beyond localhost.
- Distinctive: line 14
Before saving each starter job, show its editable prompt, provider, model, dispatch tier, recurrence, workspace behavior, and any command …
- Distinctive: line 15
Leave automatic dispatch disabled until the user has reviewed profiles and queued jobs, and ask before enabling it.
- Distinctive: line 16
Never make publishing, deployment, merge, or destructive cleanup part of a starter job without explicit approval.
- Distinctive: line 17
Prefer isolated worktree or DevX workspaces.
- Distinctive: line 18
Treat profile commands and lifecycle hooks as trusted local code.
- Very specific: line 19
Use queue and profile mutation tools only when the user asks to schedule or manage background work.
≈ “Common” is an estimate: at least 1 in 5 of the 3,727 llms.txt files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
# Redline > A local, budget-aware job dispatcher that runs queued coding-agent jobs in Claude Code, Codex CLI, Pi, Hermes, or a custom command when Codex or Claude subscription allowance is running behind pace or would otherwise expire. Redline watches 5-hour and weekly subscription allowances locally, explains each scheduling evaluation as RUN, WAIT, or fail-closed UNKNOWN, and dispatches eligible jobs, preferably into isolated Git worktrees or DevX sessions (Hermes runtimes and explicitly approved existing directories are also supported). It runs as a loopback service with a SQLite database and your existing CLI logins: no cloud service, no account, and no outbound product telemetry. It ships a local stdio MCP server (`redline mcp`) so agents can check allowance, inspect the queue, and add jobs. Agent safety contract: - Read and follow the user's repository instructions first. - For the macOS app, install only the official signed and notarized release and verify its SHA-256 checksum. Never bypass Gatekeeper, disable quarantine, or install an unsigned build. - Ask before replacing an existing configuration, and before installing or authenticating another harness or tool. - Never request, print, or store account passwords or app-specific passwords. - Never expose the loopback service beyond localhost. - Before saving each starter job, show its editable prompt, provider, model, dispatch tier, recurrence, workspace behavior, and any command it could execute, and get the user's approval. - Leave automatic dispatch disabled until the user has reviewed profiles and queued jobs, and ask before enabling it. - Never make publishing, deployment, merge, or destructive cleanup part of a starter job without explicit approval. - Prefer isolated worktree or DevX workspaces; use an existing-directory workspace for code-changing jobs only when the user explicitly accepts direct edits. - Treat profile commands and lifecycle hooks as trusted local code; change them only at the user's request. - Use queue and profile mutation tools only when the user asks to schedule or manage background work. To check eligibility, call `redline_scheduler_evaluate` (it records a decision but never launches work), not `redline_scheduler_dispatch`. ## Essential documentation - [README](https://github.com/croutoncreations/redline/blob/main/README.md): Purpose, installation, how scheduling works, MCP setup, and screenshots. - [Getting started](https://github.com/croutoncreations/redline/blob/main/docs/getting-started.md): Install the app, confirm usage monitoring, create a profile, and queue a first job. - [Scheduling and the allowance model](https://github.com/croutoncreations/redline/blob/main/docs/scheduling.md): Policies, pace gap, reserve, dispatch tiers, firesale, effective reset horizons, banked-reset redemption, and calibration. - [Execution profiles and tasks](https://github.com/croutoncreations/redline/blob/main/docs/profiles-and-tasks.md): Harness, model, and workspace configuration and the task lifecycle. - [Troubleshooting](https://github.com/croutoncreations/redline/blob/main/docs/troubleshooting.md): Common failures and corrective checks. ## Agent tasks - [Agent-assisted install](https://github.com/croutoncreations/redline/blob/main/docs/agent-install.md): Copyable install prompt and what an agent must not do. - [MCP and agent access](https://github.com/croutoncreations/redline/blob/main/docs/mcp.md): Server setup for Claude Code, Codex CLI, and Pi bridges, the read-only and state-changing tool reference, and a suggested agent instruction. - [CLI reference](https://github.com/croutoncreations/redline/blob/main/docs/cli.md): Commands for status, queueing tasks, and run inspection. - [HTTP API](https://github.com/croutoncreations/redline/blob/main/docs/api.md): The authenticated loopback API used by the dashboard, CLI, and MCP server. ## Optional - [Runs, logs, and notifications](https://github.com/croutoncreations/redline/blob/main/docs/runs-and-notifications.md): Run results, artifacts, bounded logs, banked-reset reminders and redemption notifications, and firesale start/end events. - [Hermes integration](https://github.com/croutoncreations/redline/blob/main/docs/hermes.md): Dispatching jobs to Hermes runtimes. - [Native macOS app](https://github.com/croutoncreations/redline/blob/main/docs/native-macos.md): Menu-bar app, bundled service and CLI, and release packaging. - [Run as a launchd service](https://github.com/croutoncreations/redline/blob/main/docs/launchd.md): Headless macOS service setup. - [Mobile dashboard](https://github.com/croutoncreations/redline/blob/main/docs/mobile.md): Viewing Redline from a phone. - [Architecture](https://github.com/croutoncreations/redline/blob/main/docs/architecture.md): Service, scheduler, collectors, and storage design. - [Changelog](https://github.com/croutoncreations/redline/blob/main/CHANGELOG.md): Release history.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

