# Are granular tool permissions a good approach for coding agents? — retained record

> Retained reddit record from reddit:AI_Agents.

Canonical: https://highsignal.app/data/records/7e106493370ad8e3

[← sources](https://highsignal.app/data)


source: [reddit](https://highsignal.app/data/reddit) (reddit:AI_Agents)  published 2026-10-20 16:44 ingested 2026-10-21 02:35

[original source ↗](https://www.reddit.com/r/AI_Agents/comments/1xax910/are_granular_tool_permissions_a_good_approach_for/)

## retained text



```
Hi, I've been thinking about how granular permissions for coding agents should be.

For example, an agent might need access to a GitHub tool, but should it be allowed to interact with every repository? Or should you be able to restrict what arguments it passes to that tool (let's not assume the underlying credential is scoped to the resources, many tools do not have scopes for credentials)?

We recently shipped MCP argument-level rules and JSON body rules for API requests, so we can enforce permissions beyond just allowing or blocking individual tools and endpoints. You can still allow/deny tools, this is like more gradual option. It's part of our open source CLI for coding agents.

I'm curious what you about this. Is this something you'd actually use in your agent workflows?
```



document fetched 2026-10-21 00:18 · canonical https://reddit.com/r/AI_Agents/comments/1xax910/are_granular_tool_permissions_a_good_approach_for
