anthropics / anthropics/claude-code
[FEATURE] Batch tool use approval as a call-scoped token
- 主要语言
- Python
- 星标
- 145k
- 派生
- 23.1k
- PR 合并指标
- PR 指标待抓取
描述
### Preflight Checklist
- [x] I have searched [existing requests](https://github.com/anthropics/claude-code/issues?q=is%3Aissue%20label%3Aenhancement) and this feature hasn't been requested yet
- [x] This is a single feature request (not multiple features)
### Problem Statement
Right now, approvals are per-call. There's no mechanism for Claude to say "here are the 8 Zoom reads I'll need, approve them all" and have that serve as an authorization token for those specific calls. Each call triggers its own prompt regardless. For example, Zoom already has trusted credentials at the MCP level. The per-call Claude Code prompt is a second, separate gate that can only be bypassed via allowedTools in settings.json, so if you're a part of an org that allows the connector but restricts auto approval, any recurring tool use become sitting there with Claude open to make sure you don't miss a request and it takes all of your time. I
### Proposed Solution
If Claude could just add a toggle to the main connectors settings page that allows users to select whether to have these read/write requests batched (read batched with read, write batched with write where the workflow allows for it) and then the user could just receive a single notification that they have a batch of requests for review. These batched requests would live in the relevant chat and would not time out; the prompt session would become dormant until the user approves (reactivates session) or denies (kills session).
### Alternative Solutions
I asked claude code to build me a tool that batches these requests but said it could not.
### Priority
Critical - Blocking my work
### Feature Category
CLI commands and flags
### Use Case Example
_No response_
### Additional Context
_No response_
贡献指南
这个仓库没有索引到贡献指南
调研方向
Start with the connector settings behavior and settings.json references in the issue, then trace where per-call approval prompts and allowedTools bypasses are handled. Define how read and write requests are grouped, how approval or denial affects the dormant session, and verify the behavior through the project's existing approval tests.
由索引模型根据 Issue 内容生成。
评估
- 技术栈
- python
- 领域
- authorization, cli
- Issue 类型
- 功能
- 难度
- 5/5
- 预计耗时
- 一周以上
- 活跃度
- 活跃
- 描述清晰度
- 基本清楚
- 新手友好度
- 35/100