Who should use downscope?

downscope is an open-source project built for developers.

Yes — downscope is open source under the MIT license and free to use.

What platforms does downscope run on?

downscope runs on the command line.

Is downscope still active?

The GitHub repository shows 6 commits in the last 90 days.

What are alternatives to downscope?

Similar projects tracked by PulseGate include constraintloop, trajectory-judge, and COHESION.

downscope is developed by Bogdan Colceriu.

How long has downscope been around?

downscope first shipped in 2026.

Back to the index

downscope

Name: downscope
Author: Bogdan Colceriu

PyPIInfrastructure

downscope is a least-privilege audit tool for Claude Code that analyzes which permissions you actually use, which are dead weight, and which are dangerous. It derives insights from your real session history to help secure AI agent deployments. Aimed at developers and security-conscious teams working with Claude Code and similar AI coding agents.

Open SourceMITCLI

Visit PyPI

Overview

4 features

In the LLM eval & observability space, downscope takes a focused approach. It focuses on identifying which AI agent permissions are actually used, unused, or potentially dangerous from real usage data. It is built as an open-source project for developers. downscope is open source under the MIT license. It ships for the command line.

Bogdan Colceriu builds and maintains downscope, and it first shipped in 2026. The project is developed in the open on GitHub with 6 commits in the last 90 days. Among its 4 catalogued features are permission audit, least-privilege analysis, and session history review.

Summary written by a language model from the project’s public pages.

Permission audit
Least-privilege analysis
Session history review
Security analysis

Built with & integrations

AI providers

anthropic

Runs on

CLI

Trust & compliance

License

MIT

Verified signals

HTTPSOpen Source GitHubActive maintenance

Indexing history

What PulseGate has recorded for this listing

Indexed5 Aug · 21:20 UTC
downscope verified against its public source
Source: PulseGate · Open

Frequently asked questions about downscope

What is downscope?: Downscope focuses on identifying which AI agent permissions are actually used, unused, or potentially dangerous from real usage data. It is catalogued under LLM eval & observability on PulseGate.
Who should use downscope?: downscope is an open-source project built for developers.
Is downscope free?: Yes — downscope is open source under the MIT license and free to use.
What platforms does downscope run on?: downscope runs on the command line.
Is downscope still active?: The GitHub repository shows 6 commits in the last 90 days.
What are alternatives to downscope?: Similar projects tracked by PulseGate include constraintloop, trajectory-judge, and COHESION.constraintloop trajectory-judge COHESION
Who makes downscope?: downscope is developed by Bogdan Colceriu.
How long has downscope been around?: downscope first shipped in 2026.

At a glance

Platforms: Cli
Languages: English
Open source: Yes (GitHub)
License: MIT
First seen: 5 Aug 2026
Built for: developers
Model: Open source
Solves: Identifying which AI agent permissions are actually used, unused, or potentially dangerous from real usage data.

Registered as

GitHub: bogdancolceriu/downscope
PyPI: downscope

Developer

Bogdan Colceriu

Solo developer

GitHub

Open source

View on GitHub

Stars: 0
Forks: 0
Open issues: 0
Last commit: 5 Aug 2026
Commits 90d: 6
Contributors: 1
Authorship: Solo
Default branch: main

Live coverage

Identity confidence: Low · 64
Indexed: 5 Aug 2026
Lifecycle: Alive
First seen: Aug 2026
Last seen: 5 Aug 2026

Identity audit (13)

Slug: downscope-pypi-org
Lifecycle state recorded: 5 Aug 2026
Verification state: Indexed for public listing
Listing state: Listed: yes
Index status: Included in index
Latest evidence snapshot: 5 Aug 2026
Timeline basis: Indexed-at chronology (no inferred launch/funding milestones).
Name from: Written by a language model from the project's public pages.
Category from: Assigned by a language model.
Summary from: Written by a language model from public pages.
Languages from: Detected by a language model from page content.
Last updated: 5 Aug 2026
Canonical URL: https://pypi.org/project/downscope

Ship this? Send a correction — no account, and you get a link to follow it.

Also in LLM eval & observability

Same category — not a similarity match