Blume

blume is a narrow, always-on desktop companion designed to monitor and improve coding agents. It sits alongside popular agent harnesses such as Codex, Claude, and Cursor, watching how each agent operates and surfacing real-time activity through a clear interface. Instead of requiring developers to manually supervise every run, blume provides a unified view of agent tasks, branches, and statuses, allowing teams to spot issues and understand behavior at a glance. blume goes beyond basic oversight by actively helping to keep agent configuration healthy. It observes when chat instructions or unstructured feedback contradict the existing rulebook, then prompts users to tidy up rules, skills, hooks, and corrections. This keeps the knowledge base in sync with actual behavior and reduces the need for repetitive steering. The tool is built as a lightweight local desktop application, keeping data on the user's machine and providing low-latency updates. The product is positioned for teams that rely on coding agents across different harnesses and need a systematic way to maintain their evolving playbooks. It appears to be developed by blume.codes and is likely aimed at developer teams who have adopted AI-assisted coding workflows.

Category: AI Agent
Pricing: Unknown

Problem solved

Developers lose time manually supervising coding agents and maintaining scattered rules and instructions. blume solves this by offering a single desktop interface that monitors agent activity, surfaces contradictions between chat feedback and established rules, and helps users iteratively refine the agent's rulebook to improve performance and consistency.

Use cases

Monitor coding agent tasks across multiple harnesses in real time
Detect when agent chat or feedback conflicts with configured rules
Review and approve or improve agent behavior from a single dashboard
Maintain a clean repository of rules, skills, hooks, and corrections
Track progress of agent assignments and feature branches
Align agents with project standards through structured feedback

Audience

Developers
Enterprise
Startups

Pros

Provides a unified view of activity across different coding agent harnesses
Detects conflicts between ad-hoc instructions and formal rules
Runs locally and always-on, minimizing overhead
Helps keep agent rules, skills, and hooks continuously maintained

Cons

Requires initial setup to connect with supported agent harnesses
Currently supports a limited set of harnesses (Codex, Claude, Cursor)
No clear indication of pricing or open-source availability

Official website: https://blume.codes/

Canonical LambdaBase page: https://www.lambdabase.com/tools/blume

← Directory

Blume

Unknown

• Updated Sep 8, 2026, 03:29 AM

blume is a narrow, always-on desktop companion designed to monitor and improve coding agents.

Use Cases ▼

Monitor coding agent tasks across multiple harnesses in real time

Detect when agent chat or feedback conflicts with configured rules

Review and approve or improve agent behavior from a single dashboard

Maintain a clean repository of rules, skills, hooks, and corrections

Track progress of agent assignments and feature branches

Align agents with project standards through structured feedback

What Problem It Solves ▼

Key Features ▼

✓

Provides a unified view of activity across different coding agent harnesses

✓

Detects conflicts between ad-hoc instructions and formal rules

✓

Runs locally and always-on, minimizing overhead

✓

Helps keep agent rules, skills, and hooks continuously maintained

Overview

Full Description ▼

blume goes beyond basic oversight by actively helping to keep agent configuration healthy. It observes when chat instructions or unstructured feedback contradict the existing rulebook, then prompts users to tidy up rules, skills, hooks, and corrections. This keeps the knowledge base in sync with actual behavior and reduces the need for repetitive steering. The tool is built as a lightweight local desktop application, keeping data on the user's machine and providing low-latency updates.

The product is positioned for teams that rely on coding agents across different harnesses and need a systematic way to maintain their evolving playbooks. It appears to be developed by blume.codes and is likely aimed at developer teams who have adopted AI-assisted coding workflows.