Niches · Cloud & infrastructure
Change, resize or clean up live resources
No company's docs show a full fit yet. The listings below do the job with a limit, or not at all.
What counts as done
Only the named resource in the named environment changes; the tool shows the change before making it (dry run or plan), a destructive step waits for a person's approval, and the result is read back.
How agents ask for it: resize the database volume
, clean up unused resources
, scale this service
, delete the old staging environment
Does the job, with a limit 9
It does the job, but something in the docs or a user's account will get in an agent's way. The limit is named in each row.
| Listing | Why |
|---|---|
| AWS MCP Servers C Amazon Web Services, Inc. | REQUIRE_MUTATION_CONSENT asks for consent before writes and IAM limits scope, but both safety flags default to false and there is no plan or dry-run summary of a change. source |
| Azure MCP Server BB Microsoft Corporation | Writes run under the caller's RBAC with no confirmation step; the README warns that autonomous clients may perform destructive actions, and there is no dry run. source |
| Cloudflare CLI (cf) B Cloudflare, Inc. | --dry-run prints the request as JSON, and in non-interactive sessions destructive commands abort without --force; but the abort exits 0, and the agent can pass --force itself, so no person approves. source |
| Cloudflare MCP Servers BB Cloudflare, Inc. | Writes need a scope that the client is asked to add on insufficient_scope, but once granted there is no per-call confirmation or dry run. source |
| DigitalOcean MCP Server B DigitalOcean, LLC | Custom-scope tokens can narrow what a token may touch, but delete tools such as apps-delete run with no confirmation step or dry run. source |
| Netlify API + MCP BB Netlify, Inc. | Draft deploys preview a change before publishing, but tokens can't be narrowed to one site or environment and deletes have no confirmation step. source |
| Railway MCP Server C Railway Corporation | The local server returns a preview and requires confirm: true for destructive tools, but tokens carry no per-resource scope: acc-01's domain token deleted a volume shared with production, and acc-02 reports Railway's agent wiping a database. source |
| Render MCP Server B Render Services, Inc. | Changes are limited to triggering deploys and updating environment variables, with no deletes, but there is no preview of a change and no per-environment credential. source |
| Terraform MCP Server BB HashiCorp, Inc. (an IBM Company) | plan_only and get_plan_details show the plan before action_run applies it, and applies, deletes and force unlock need ENABLE_TF_OPERATIONS; but create_workspace and update_workspace, marked destructive, are not behind that flag. source |
What stops agents finishing
- any token can delete any resource, whatever it was made for
- staging and production share a resource the agent cannot see
- no confirmation step on delete once a write scope is granted
Around it
Before: infrastructure.inspect-debug. After: observability.verify-deploy.
Where this comes from
Weight 14 of 100 in Cloud & infrastructure: 5 of 32 accounts, four of them production or staging deletions (PocketOS on Railway, the Terraform destroy on AWS); the strongest failure evidence in the category
The same as data: /api/v1/use-cases/infrastructure.change-live-resources.json, or the MCP tool get_use_case with id infrastructure.change-live-resources. Other jobs in this category: /api/v1/niches?category=infrastructure.
For companies
Do agents find, use and choose your tools?
An agent-readiness audit runs our probes, task suite and eight reviewer agents against your public and internal tools, and comes back with a scorecard, the transcripts of what failed, and a fix list in priority order. From $2,500, re-run included. We never take payment to move a rank. We do help companies earn one.