Claude Code 2.1.289 plugin loading fix: recheck Sume MCP after upgrade
Claude Code 2.1.289 fixed plugin loading after an upgrade. A four-call read-only check confirms the Sume MCP connection and scopes before you run a paid job.

After upgrading Claude Code, confirm Sume with mcp_health and tools_list before any paid call. Version 2.1.289 (Oct 3, 2026) fixed plugin loading after an upgrade, according to the Claude Code changelog, but a read-only check costs nothing and tells you the real state.
This is the short sequence Sume's docs recommend for verifying a new connection.
Reconnect if needed
If the server is missing after the upgrade, add it again and sign in. These are the two commands from the Sume quickstart.
claude mcp add --transport http sume https://mcp.sume.com/mcp
claude mcp login sumeThe four-call check
Ask the agent to call mcp_health first and confirm that authenticated.auth_source is mcp_oauth when you connected with OAuth.
| Tool | What it confirms |
|---|---|
| mcp_health | Endpoint, auth source and safety posture. |
| tools_list | Every tool visible to this session, with safety metadata. |
| account_me | The workspace account context. |
| balance_get | Spendable balance before you submit anything. |
Read the tool list as a scope test
The tool list is also the quickest way to see which scope you hold. Default hosted OAuth grants mcp:read only, so mutating and paid tools stay hidden. If you expect to run generate_image or jobs_cancel and they are missing, the session probably lacks mcp:write.
With read-only scope, calling a write or paid tool by name returns insufficient_scope instead of running. Grant Write on the consent page, use an API-key session, or use the Developer API or CLI for those calls.
Before the first paid call
Once the connection is verified, do one more free step before real spend.
- Call
tools_schemawithnameset to the tool you plan to use, such asgenerate_image. - Call the tool with
dry_run=true, or callgeneration_admission_preview, and read the estimate. - Submit with a fresh
idempotency_key. Addmax_spend_usdif you want a cap. - Wait with
jobs_waitin slices of at most 55 seconds, then readjobs_result.
If jobs were running during the upgrade
Restarting a client does not stop a job that Sume has accepted. Jobs are durable, and reading jobs_list or jobs_status with the ids you stored shows where each one stands. Do not resubmit a paid request because your session was interrupted.
Sources
Related posts
More in Developers
- Claude Code's 60 s MCP timeout: Sume sync waits 30 s, video polls
Claude Code 2.1.287 caps MCP tool calls at 60 seconds per server. Sume's 30 s sync wait and 55 s jobs_wait fit under it; a video render needs repeated waits.
- Claude Files API is GA: review Sume outputs by job id and media URL
The Claude Files API left beta on Aug 19. When Claude reviews generated media, store the Sume job id and media.sume.com URL, not raw provider links.
- Cloudflare AI Search bills from Nov 1: split retrieval from renders
Cloudflare's October 1 changelog makes AI Search GA with usage billing from November 1, 2026. How to keep retrieval costs separate from Sume render costs.
- Codex 0.160 agent history 'Show more': recover Sume jobs by job list
Codex CLI 0.160.0 adds Show more pagination to agent command center history. If a thread scrolls away, Sume's GET /v1/jobs list still holds every job.
Written by Sume