Authentication & Scopes

How MCP tokens, access levels, and permissions work.

Tokens

Every request to the MCP server must include a bearer token:

1Authorization: Bearer jr_mcp_...

Tokens are minted in Settings > MCP Server by workspace owners and admins. They are:

  • Workspace-scoped — a token only ever sees and touches the workspace it was created in
  • Hashed at rest — JotReview stores only a hash; the raw token is shown once at creation
  • Revocable — revoke a token at any time and it stops working immediately
  • Auditable — every change an agent makes is recorded in your activity feed, tagged as coming from MCP

Access levels

When you create a token you pick one of two access levels:

Access levelRoleWhat it can do
Read onlyViewerList and read requests, roadmap, changelog, comments, tags, and boards. No changes.
Read & writeAdminEverything above, plus create, update, moderate, and delete.

Choose Read only for reporting or analysis agents, and Read & write for agents that manage your backlog.

Scopes & roles

Under the hood, each token carries a set of scopes and a role ceiling. Every tool declares the scope and minimum role it needs, and the server denies the call if the token falls short.

ScopeGrants
requests:readList and read requests and the roadmap
requests:writeCreate requests and change status, tags, or board
requests:moderateApprove or reject requests
changelog:readRead changelog entries
changelog:writeCreate, update, publish, and delete entries
comments:readRead comments
comments:writeAdd comments and replies
tags:writeCreate, rename, and delete tags
boards:writeCreate, rename, and delete boards
workspace:readRead workspace info

A read-only token receives all the read scopes; a read & write token receives them all.

Security best practices

  • One token per agent — separate tokens let you revoke a single client without disrupting others
  • Prefer least privilege — give reporting agents a read-only token
  • Rotate periodically — revoke and recreate long-lived tokens on a schedule
  • Watch the activity feed — agent-driven changes appear there, tagged as MCP and attributed to the token's creator