HOW IT WORKS

RunTrim controls the run around your agent.

Start with one command. RunTrim scopes the work before the agent starts, watches what changes, and preserves the next safe step.

If you are unsure, run runtrim start.

Install CLI Back to homepage

Step 1

Go

runtrim go "your task"

Prepare a guarded prompt, copy it for your agent, and save the run locally.

Step 2

Paste into your agent

Use Claude, Codex, Cursor, ChatGPT, or your configured agent.

Step 3

Monitor

runtrim panel --monitor

Keep local state visible while the agent works.

Step 4

Check

runtrim check

Review changed files, proof gaps, and drift before continuing.

Step 5

Continue

runtrim continue --reason usage_limit

Create a continuation prompt when context, usage, or the session runs out.

What local memory means

RunTrim stores project memory in .runtrim so the next session does not restart from scratch.

Guarded prompts
Run status
Changed file paths
Protected systems
Verification debt
Next safe action
Continuation prompts

What RunTrim reads

Task text
Project metadata
Package and framework signals
Local .runtrim memory
Git diff file paths

What RunTrim does not upload in V1

Source code
.env files
Secrets
Raw file contents

RunTrim does not need to upload source code to create a guarded run contract. Cloud sync, when enabled, is metadata-only and currently part of Pro early access.

Start with the local CLI.

Free local CLI works without an account. Cloud memory is rolling out through Pro early access.

npm install -g runtrim
runtrim go "your task"

Install CLI