Tools & Permissions

When you connect Claude Code to a project, the gateway uses your permissions on that project — the same access the in-product project agent has for you. There is no separate gateway role to pick. Which tools appear follows your org role (or the org’s project agent-profile default). See Members & Roles.

​

What your permissions unlock

Project tasks and memory are always available. Everything else follows the matching permission on your role.

​

The tool groups

​

Project tasks

Read and manage the project’s task list — the same persistent work items you see on the project dashboard. Claude can list tasks, create new ones, update status, and remove them.

​

Project memory

Read and write the project’s persistent memory: notes and context that survive across sessions, scoped to this project.

​

Data store

List the project’s data stores, inspect their schema, run queries, and generate TypeScript types from a store’s live schema so the types in your checkout match the real database. With data store read access, Claude also gets write tools over the gateway — the same as the in-product project agent: schema changes, data mutations, and creating a data-store branch (an independent copy of the store’s tables and rows; see Branches and environment routing) to experiment without touching another branch’s data. Writes still target the connection’s configured agent branch.

​

Cross-app

Call endpoints on your other apps through the platform’s cross-app communication. With skills execution access, the project gateway exposes both GET and non-GET calls — data store write access is not an additional requirement. This differs from the org-level AI Integrations gateway, where OAuth read vs read & write scope controls GET vs non-GET.

​

Blueprint updates

When your role grants repo access, Claude can check whether this Blueprint-derived app has a newer published version (get_blueprint_update) and record a completed merge (complete_blueprint_update). The tools prepare an exact fetch ref, walk through a local git merge, and verify the result on the server before updating the baseline. See Update from a Blueprint for the end-to-end workflow.

​

Skills

Discover Stardeck skills — focused guides for building on the platform. Claude can load design skills (UI, dashboards, i18n, and more) as content directly, and for SDK skills (auth, email, payments, data store, …) it gets a pointer to read that package’s SKILL.md from your checkout’s node_modules. Available with skills execution access.

​

App roles & permissions

Read and configure an app’s full roles and permissions model: deployment permissions, deployment roles, role-permission assignments, default sign-up role, organization-member access policy, and which deployment permissions each organization role holds for this app. Read access requires the app roles & permissions read permission; writes additionally require the write permission.

​

Branches and production data

Data store writes resolve to the connection’s configured agent branch — the same branch the in-product agent uses, not necessarily a throwaway one.

​

Security model

The connection has two independent gates:

• Who can connect — the person authorizing must be a member of the organization that owns the project. A token issued for one organization can’t reach another organization’s projects.
• What Claude can do — your permissions on the project (same resolution as the project agent). The token decides who connects; your role decides what they can do once connected.

On top of that:

• Off by default — the gateway is disabled per project until you turn it on.
• No stored secret — authorization happens in your browser; the .mcp.json in your repo holds only the project URL.
• Audited — every call through the gateway is logged and tagged so it’s distinguishable from in-product agent activity.
• Client-side prompts — Claude Code’s own permission prompts add a final confirmation before tools run on your machine.

​

Next steps