MCP Reference

Tap & Scale exposes the same reads and money-safe writes as the app through a Model Context Protocol server. Add it to an MCP client and your agent can audit, plan, and optimize an account under the same safe-apply policy — bounded changes auto-apply, the rest queue for your approval.

The server speaks Streamable HTTP at /api/mcp. For callers that aren't MCP clients — cron jobs, backend services, curl — the same capabilities are available over the REST API. Both transports run on one tool set, so they never drift. MCP authenticates over OAuth 2.1 with no token to mint, or — for a headless agent — with the same tsk_ key the REST API takes, verified by the same code, so revoking a key stops both surfaces.

How tools behaveCopy Anchor Link

• Every tool returns { ok: true, ... } or { ok: false, error }. Read the error — it usually names the next step.
• Reads are safe. asa_propose_* tools move real money and run through the safe-apply policy.
• A connected session reaches the whole workspace, so call asa_list_ad_accounts first and pass the account's projectId to every account-scoped tool. Workspace-level tools take none.
• Apple's vocabulary applies throughout: taps, installs, CPT, CPA, storefronts, EXACT|BROAD, ACTIVE|PAUSED. See vocabulary.

Tool catalogCopy Anchor Link

Accounts & setupCopy Anchor Link

Reads & analysisCopy Anchor Link

Growth & keyword ideationCopy Anchor Link

Writes (money-moving propose_*)Copy Anchor Link

Approvals (operator-only, over MCP)Copy Anchor Link