How do I pass an Authorization header to an MCP server in Continue?

Put it under `requestOptions.headers` on the `mcpServers` entry. `requestOptions` applies to `sse` and `streamable-http` servers and uses the same format as model request options.

Can I keep the Comma token out of Continue's config file?

Yes — reference it as `${{ secrets.COMMA_API_TOKEN }}` and store the value in Continue's secrets, so the YAML is safe to commit.

Do I need to run anything locally for Comma's MCP server?

No. Comma's MCP server is hosted at https://commareports.com/api/mcp over Streamable HTTP. The config entry and a `comma_sk_` token are the whole setup — unless you use the `mcp-remote` bridge, which runs via `npx`.

Is Comma free to use from Continue?

Yes. Publishing reports, inviting commenters and keeping revisions are free. Routines — scheduled runs Comma hosts — are the paid part.

Use Comma with Continue

Continue's agent mode can call MCP tools from any assistant config. Comma is a hosted server, so the whole setup is a YAML block with a URL and a header — and the token can come from Continue's secrets rather than sitting in the file.

Add Comma to Continue

Add a block to mcpServers in your config.yaml, or drop a standalone file in .continue/mcpServers/ in the workspace. Continue also accepts a JSON file in Cursor/Cline format at .continue/mcpServers/mcp.json.

mcpServers:
  - name: comma
    type: streamable-http
    url: https://commareports.com/api/mcp
    requestOptions:
      headers:
        Authorization: Bearer ${{ secrets.COMMA_API_TOKEN }}

Replace comma_sk_… with a token from Settings → API tokens → New token on Comma (or export it as COMMA_API_TOKEN where the snippet reads an env var). The default scopes — reports:write, comments:read, comments:write — cover publishing and review.

Continue gotchas

Headers go under requestOptions. That's where Continue reads headers for streamable-http servers — if Comma answers 401, the header is in the wrong place or missing Bearer .
Use ${{ secrets.COMMA_API_TOKEN }} so the config can be shared or committed without the token in it.
MCP tools are only offered in Agent mode — in Chat mode the model won't see create_report.

What the agent can do

Capability	Tools
Publish a report	`create_report`, `update_report`, `create_report_from_url`
Find existing reports	`list_reports`, `search_reports`, `get_report`, `read_report`
Leave / reply to comments	`add_comment`, `reply_to_comment`, `set_comment_status`, `list_comments`
Ask for review, share	`request_review`, `set_report_sharing`, `get_report_signoffs`
Manage assets	`upload_report_asset`, `list_report_assets`, `delete_report_asset`

Continue picks the tool from the request — there are no Continue-specific commands to learn.

A workflow that sticks

Ask Continue to write up the work as a self-contained HTML report and publish it to Comma. It calls create_report and returns a link.
Send the link. Reviewers open a rendered page and leave comments anchored to the exact line, cell or chart they mean — no account needed to read a public link.
Next session, ask Continue to address the comments. It reads the threads with list_comments, ships a revision with update_report (same URL), replies, and resolves them.

Reports belong to your workspace, not to the agent. A report published from Continue can be revised tomorrow from Claude Code or Cursor, and the link people bookmarked keeps working.

Token shape

Referenced from Continue's secrets, so config.yaml itself carries no credential.
Scoped — you choose which actions it can take when you create it.
Revocable in one click from Settings → API tokens; revocation applies immediately to the REST API, the MCP server and any routines the token created.
Rate-limited per token.

Details in API tokens → and Scoped tokens for AI agents →.

Try it

Add the entry, generate a token, restart Continue, and ask for a report on something you worked on today. Then leave a comment on it and ask the agent what it sees.

Generate a token →