Sume MCP config keys in Cline, Zed, Gemini CLI and Zapier
One Sume MCP URL, four spellings: streamableHttp in Cline, url in Zed, httpUrl in Gemini CLI, and a transport field in Zapier. A side-by-side table.

Every client points at the same address, https://mcp.sume.com/mcp, but each names the field differently: Cline uses type streamableHttp plus url, Zed uses url under context_servers, Gemini CLI uses httpUrl (its url key means SSE), and Zapier's MCP Client form asks for a server URL and a transport. This page lines them up so you can copy the right key.
What does each client call the field?
The table is built from each vendor's own page, read on the date in the caption.
| Client | Where | URL key | Key header |
|---|---|---|---|
| Cline CLI | ~/.cline/data/settings/cline_mcp_settings.json | url with type streamableHttp | headers.Authorization |
| Zed | settings, context_servers | url | headers.Authorization |
| Gemini CLI | mcpServers | httpUrl (url is SSE) | headers.Authorization |
| Zapier MCP Client | Apps, Add connection | Server URL form field | Bearer token field |
What happens when you send no key?
Zed's docs say that when no Authorization header is configured, Zed prompts you to authenticate through the standard MCP OAuth flow. Sume supports that flow: it answers with protected-resource metadata, sends you to its consent page on the MCP host, and issues a token. The default is read-only (mcp:read); the Write toggle adds mcp:write.
Gemini CLI defaults to dynamic discovery for its OAuth provider type, and Zapier has an OAuth option on the form. Cline's page documents headers and a CLI wizard for browser authorization.
Which credential suits which job?
A Sume OAuth token is not a Sume API key, and neither replaces the other. Do not paste either into chat.
- Exploring a catalog from an editor: OAuth, Write off.
- Generating media from an editor: OAuth with Write on, or a key.
- Unattended automation such as Zapier or CI: an API key.
- Anything paid: send dry_run true first, plus an idempotency_key.
What stays the same in every client?
Whatever the field name, the server behavior does not change. The URL is the production hosted endpoint, discovery tools are free, and the first call worth making is mcp_health followed by tools_list. Write and paid tools always need an idempotency_key, and dry_run previews a cost without submitting.
Do not put the key in the URL. Sume documents the Authorization bearer header and the x-api-key header for key sessions, and the vendor pages above show a header object or a bearer token field in each client. Where a client cannot hide the key in a header, use OAuth instead.
If a client reports a failed connection, compare the URL spelling character by character against https://mcp.sume.com/mcp, then check the transport field. Choosing SSE where Streamable HTTP is expected is the most common mismatch.
Before you rely on any of this in a team setting, test it once end to end with a throwaway prompt. Call mcp_health, call tools_list, and run one dry_run against a paid tool. Compare the tool count with what you expected for your credential. If a read-only OAuth session lists more than you expected, or a key session lists fewer, stop and recheck which credential the client is actually sending. Sume's mcp_health response reports the auth source, which settles the question quickly without guessing.
Sources
Related posts
More in Integrations
- Zapier Next Gen Zaps: pause for a human reply before a Sume agent run
Next Gen Zaps can pause for a human reply. Use that pause to approve a Sume agent run, then send the approved generation_spend_cap_usd and an idempotency key.
- Zed tool permissions for Sume: mcp:server:tool keys
Zed asks before each MCP tool by default. How to use its mcp:<server>:<tool> permission keys so Sume's reads run freely and its paid tools stay on confirm.
- Cinema 4D MCP plus Sume MCP in one Claude Desktop chat
Cinema 4D 2026.4 ships a local MCP server. Pair it with Sume's hosted MCP: render the scene locally, then animate and finish the still in Sume.
- Claude Code 2.1.294 prompt hook fix: spend limits need a command hook
Claude Code 2.1.294 fixed prompt and agent hooks written as instructions that let blocked calls through. For Sume spend limits, a command hook is still exact.
Written by Sume