[FEATURE] Allow configuring network allowlist for remote scheduled agents (routines)

Resolved 💬 3 comments Opened Apr 17, 2026 by justinp320 Closed Apr 21, 2026

Preflight Checklist

  • [x] I have searched existing requests and this feature hasn't been requested yet
  • [x] This is a single feature request (not multiple features)

Problem Statement

Remote scheduled agents (routines/triggers) run in a sandboxed environment that blocks outbound HTTP requests to any domain not on a preset allowlist. There is no self-serve way to add domains to this allowlist.

This prevents scheduled agents from calling external APIs like Slack, Linear, Notion, or any other third-party service — which significantly limits what automation pipelines can do.

We're building an automated pipeline that runs weekly to pull new messages from Slack channels via the Slack API, synthesize and categorize them into a knowledge base in our repo, then commit and push the updates. The pipeline works perfectly when run locally, but fails in the remote environment because slack.com / api.slack.com are not on the network allowlist (returns 403 Host not in allowlist).

The only current workaround is running the pipeline locally, which defeats the purpose of the automation.

Proposed Solution

Allow users to configure an allowedDomains list for their remote environments — similar to the existing sandbox.network.allowedDomains setting available for local sandboxing — so scheduled agents can reach the external services they need.

This could live in the environment configuration at https://claude.ai/code/scheduled, or as a field on the trigger/routine definition itself.

Alternative Solutions

_No response_

Priority

High - Significant impact on productivity

Feature Category

CLI commands and flags

Use Case Example

_No response_

Additional Context

Our team maintains a product wiki — a knowledge base of product decisions sourced from Slack conversations. We have a scheduled agent (routine) that runs every Monday to:

  1. Call the Slack API to pull new messages from internal channels since the last sync
  2. Filter out noise (bot messages, system messages, unanswered threads)
  3. Categorize conversations by product area (invoicing, forms, scheduling, etc.)
  4. Format and append entries to the appropriate wiki pages in our GitHub repo
  5. Update index files and an append-only operation log
  6. Commit and push the changes

This pipeline has been tested end-to-end and works locally via Claude Code. When run as a scheduled remote agent, every outbound request to api.slack.com returns a 403 (Host not in allowlist), preventing the agent from performing its core task. Without external API access, scheduled agents are limited to operations that only touch the git repo itself — which rules out a wide range of practical automation use cases involving third-party services

View original on GitHub ↗

This issue has 3 comments on GitHub. Read the full discussion on GitHub ↗