# Fix list: Coinbase MCP server (Coinbase for Agents) From Anchor Terminal's listing at https://www.anchorterminal.com/tools/coinbase-mcp, the October 2026 research run, assessed 24 October 2026. Grade C, 60.2 out of 100. This is everything the published grade says the listing lacks, the biggest possible gain to the total first. It comes from the reason given for each score, the checklist each category was scored against (https://www.anchorterminal.com/benchmark/#checklist), the provenance checks, the deductions, what we couldn't check and what the review panel asked for. A fix counts at the next check, once it's public. For a coding agent working on Coinbase MCP server (Coinbase for Agents): work through the items below in the product, its docs and its public pages. Each category gives the reason for its score, with the points each checklist item earned, and the checklist itself, so the gap is the items that earned less than their points. Change the product, not the wording, and keep a note of what you changed and where it's published. ## 1. Reliability, 55 out of 100, up to 9 more on the total Why it scored 55: Graded with the hosted lines, on the remote server at agents.coinbase.com. status.coinbase.com is a Statuspage site with Advanced Trade, API, US Derivatives and Developer Platform components and 90 days of history, but no component for the agents MCP server, so it shows the trading back end and not the server agents call. Judgement call (10 of 20). Its first history page, 1 August to 24 October 2026, shows minor incidents for derivatives positions and portfolio transfers and one major, a two-minute website slowdown on 14 October, with none for Advanced Trade order entry; 26 to 31 July sits on a second page we didn't read (20). The setup guide says it specifies no numeric request quota for the remote server and that Advanced Trade limits are not the MCP limit (0). It says to honour Retry-After on 429 or back off with jitter, orders take a `client_order_id` that returns the existing order on retry, and x402 payments take an `idempotency_key` (15). No SLA found (0). Nothing labels the remote server beta or preview (10). The checklist (https://www.anchorterminal.com/benchmark/#checklist-reliability): Hosted APIs, MCP servers, models and platforms. - 20, a public status page with component history (Statuspage, Instatus, BetterStack or the vendor's own). - 0 to 30, the incident record for the last 90 days on that page. 30 for a clean record or trivial incidents only, 20 for minor incidents only, 10 for one major outage (an hour or more of a core API down, or errors across the board), 0 for several. 5 when there's no history we could read, and the note says so. - 15, rate limits documented with numbers. - 15, documented 429 or overload handling (Retry-After, backoff guidance), and idempotency keys or safe-retry guidance where writes are involved. - 10, an SLA published for any paid tier. - 10, the surface agents use is generally available, not beta or preview. Local packages, SDKs, frameworks and stdio MCP servers. - 20, installs from an official package with supported runtimes stated. - 25, a public CI and test suite, passing on the default branch. - 0 to 25, open crash or regression issues relative to activity (25 for few and handled, 0 for many, old and unanswered). - 15, semver discipline and breaking changes called out in a changelog. - 15, version 1.0 or later, or declared stable. Protocols are read from their reference implementations, the public facilitators or servers, spec stability and test vectors. ## 2. Payments & pricing, 40 out of 100, up to 7.5 more on the total Why it scored 40: Read the way coinbase-cdp-agentkit was. The server pays x402 as a buyer from the account's USDC balance and its own endpoint isn't paid per call, so the paying-as-a-buyer step (15 of 40). No price for the server itself; trades carry Advanced Trade fees, whose schedule on coinbase.com we couldn't read past a bot check, and x402 prices are advisory and shown only after sign-in (10 of 20). Connecting and reading prices and portfolios cost nothing and need no card, but anything beyond reads needs a verified and funded Coinbase account. Judgement call (15 of 20). A person must sign in to Coinbase in a browser and approve the grant, and only allowlisted clients can register (0). The checklist (https://www.anchorterminal.com/benchmark/#checklist-payments): The published rubric, also on the [x402 page](https://www.anchorterminal.com/x402/). - 40, a machine payment protocol (x402, MPP or L402) on the tool's own endpoints. 10 to 30 when it covers only some endpoints or only goes through a third party, and the note says which. - 20, per-call or per-unit pricing published without a login. 10 for public plan-only pricing, 0 for "contact sales" or prices behind a login. - 20, a free tier or trial that doesn't need a card. - 20, autonomous onboarding, meaning an agent can get access without a person signing up in a browser (keyless use, x402, a programmatic key API). Payment platforms and agent wallets rarely charge for their own API over a machine protocol, so the first line has steps for them, and the highest one that applies counts. 40 when x402, MPP or L402 runs on all their own endpoints, 30 when it runs on part of their own API, 25 when their merchants can accept one, 20 for running a facilitator, 15 for paying as a buyer, and 0 when the only protocol is their own. Merchant acceptance sits above a facilitator because the platform's own customers can charge agents through it, while a facilitator settles for sellers who wire up the protocol themselves. The counter-argument (a facilitator does more for the protocol as a whole) has a point. Each note says which step applied. Open-source software you run yourself is scored on its hosted or paid option if it has one. A free, self-hosted package with nothing to buy gets 20, 20 and 20 for the last three lines, and 0 to 40 for the first only if it ships a payment protocol. ## 3. Schema & documentation, 65 out of 100, up to 5.7 more on the total Why it scored 65: No public tool manifest. `tools/list` answered 401 without a token, so the JSON Schema of the 29 tools couldn't be read. The docs say the tool schemas carry field descriptions and enums, and the companion CLI's bundled command spec types every parameter. Judgement call (10 of 25). llms.txt, a Markdown twin of every docs page and a self-contained setup guide written for agents (10). The docs give one line per tool, and the guide has a routing matrix saying which tools to use or avoid for spot, equities and futures, but the tool descriptions as served weren't read (12 of 20). Enums and constraints are described for sessions, order types and sizes, not verified on the server (8 of 15). Order and x402 examples, and two error tables with cause and fix for 401, 403 `Missing required scopes`, `PERMISSION_DENIED`, `MISSING_FIELDS`, `INVALID_VALUE`, `INVALID_FORMAT` and x402 failures (13 of 15). Dated Coinbase for Agents entries in the CDP changelog and a versioned guide (0.2.3), but no server version is published and the October changes have no changelog entry (12 of 15). The checklist (https://www.anchorterminal.com/benchmark/#checklist-schema): APIs and MCP servers. - 25, a machine-readable contract (a public OpenAPI file or similar; for MCP, typed JSON Schema inputs on every tool). - 10, llms.txt or Markdown docs served for agents. - 0 to 20, descriptions that say what a tool is for, when to use it and when not to, read from the tool definitions in the source or the API reference. - 0 to 15, typed inputs with enums, constraints and required fields, and no free-form JSON blobs. - 0 to 15, examples and documented error responses. - 15, versioning and a public changelog. Models are read from the API reference, the OpenAPI file, llms.txt, the structured-output and tool-use docs and the model cards. Frameworks from docs a model can follow, typed interfaces, examples and the API reference. ## 4. Agent ergonomics, 66 out of 100, up to 5.5 more on the total Why it scored 66: Graded on the remote server. 29 tools per the docs, with no toolsets or read-only subset documented (15). `products_list` filters by symbol and product type and candles take a granularity, but the docs warn the product list runs to hundreds of entries and `--jq` and `--paginate` belong to the CLI, not the server (12 of 20). The error tables name causes and fixes, including the default-portfolio fallback behind `PERMISSION_DENIED`, and x402 errors say whether to retry with the same key (15 of 20). Client order IDs and x402 idempotency keys make retries safe. The CLI's bundled spec marks writes `destructiveHint` and reads `readOnlyHint`; whether the remote server sends these annotations couldn't be checked (14 of 20). Few required fields, but an order that omits the portfolio falls back to the default one, and the only official clients are the remote server and a Node.js CLI (10 of 15). The checklist (https://www.anchorterminal.com/benchmark/#checklist-ergonomics): - 0 to 25, context cost. For MCP, the number and size of the tool definitions (25 for ten or fewer compact tools, 15 for 11 to 30, 5 for more than 30, plus up to 10 back for toolsets, dynamic loading or read-only subsets). For APIs, whether responses can be sized (field selection, limits, summaries). - 20, pagination, filtering and output-size controls. - 20, actionable, documented error responses, codes and messages an agent can recover from. - 20, idempotency or safe retries, and for MCP the `readOnlyHint` and `destructiveHint` annotations. - 15, sensible defaults, few required parameters, and official SDKs in at least two languages. Models are read for tool use, structured output, prompt caching, context length, batch and SDKs. Frameworks for how much code and how many defaults a tool-calling agent with MCP needs. ## 5. Security & auth, 70 out of 100, up to 5.3 more on the total Why it scored 70: Coinbase OAuth with authorisation code and PKCE S256, `resource` on every request, DCR and CIMD limited to allowlisted clients, ten `mcp:*` scopes, a consent screen where the person picks portfolios, and revocation under Account > Security > Connections (30). A read-only grant and a funded separate portfolio limit what an agent can touch, and transfers move funds only between the person's portfolios. No per-order size limit or server-side confirmation is documented; the June blog's trade-size limits are still "soon", and the setup guide tells agents to request every scope rather than the least needed (12 of 20). x402 fetch returns third-party data from a curated catalogue only, and the 13 October changelog removes Bazaar resources whose metadata carries prompt-injection risk; no guidance on treating returned data as untrusted was found beyond citing only values the source returns (8 of 15). Orders, transfers and payments appear in the Coinbase account and connections are listed in account settings, but no per-call log for MCP was found (8 of 15). HackerOne bug bounty named in coinbase.com's security.txt, which expired on 1 June 2024; no security.txt on agents.coinbase.com; certifications not read because coinbase.com answered a bot check (12 of 20). The checklist (https://www.anchorterminal.com/benchmark/#checklist-security): - 0 to 30, the credential model. 30 for OAuth 2.1 with scopes, or scoped and revocable keys with rotation. 20 for plain revocable API keys. 10 for one all-powerful key. 10 off when a secret can travel in a URL query string as a documented option. - 0 to 20, read-only or least-privilege modes, and confirmation or approval for destructive actions. - 0 to 15, prompt-injection posture where the tool returns untrusted content (documented mitigations or guidance). A tool that returns no untrusted content gets 10. - 0 to 15, audit logs or per-call visibility for the operator. - 0 to 20, a security programme. security.txt or a disclosure policy, a bug bounty, SOC 2 or ISO 27001, advisories handled in public. Models are read for retention, whether API data trains models (and whether that's off by default), zero-retention options and certifications. Frameworks for telemetry defaults, approval hooks, guardrails and sandboxing. ## 6. Maintenance & community, 58 out of 100, up to 3.7 more on the total Why it scored 58: Read as a closed hosted service. The last dated change to the server in the CDP changelog is US stocks on 21 August 2026; the docs added three x402 tools in October (the overview of 2 October still calls them coming soon, the MCP page was modified on 9 October) and the setup guide on 22 October, with no changelog entry. Judgement call (25 of 30). Changelog entries for futures (17 August) and stocks (21 August) both name the MCP server, and the CLI built from the same agent spec shipped six versions from 14 August to 30 September (20). A public changelog, a CDP Discord for integration help and a one-way feedback tool limited to six notes a day, none tested (10 of 25). Not in the official MCP registry; a search for coinbase on 24 October found only third-party servers (0). The CLI has no runtime dependencies but runs a postinstall script and has no public repository or CI (3 of 10). The checklist (https://www.anchorterminal.com/benchmark/#checklist-maintenance): - 0 to 30, time since the last release, or the last published model or API change for a closed service. 30 within 30 days, 20 within 90, 10 within 180, 0 older. - 20, at least three releases or dated changelog entries in the last 90 days. - 0 to 25, responsiveness. Issues and pull requests answered on GitHub (the open issues and how recent the replies are). For closed services, a public changelog and a support or community channel that answers, 0 to 15. - 15, presence in the official MCP registry under a verified namespace (MCP servers), or current official SDKs (APIs and models). - 10, package health, current dependencies and CI. Models are read for deprecation notice periods and model churn rather than release counts. ## 7. Transparency & trust, 64 out of 100, up to 3.2 more on the total Made of editorial 32, provenance 95. Why it scored 64: The editorial half. A closed hosted service. The docs carry a disclaimer that the person is responsible for every trade an agent makes, but name no terms for agent access, and the Coinbase User Agreement answered a bot check and wasn't read. The CLI is MIT on npm (10 of 30). Per the coinbase-cdp-agentkit check of 1 October 2026, the privacy policy (updated 27 May 2026) names Coinbase, Inc., lists processing countries and ties retention to need without numbers; nothing specific to MCP calls or the feedback tool's screening was found (12 of 30). No deprecation policy or notice period for the server or its tools (0). Processing countries are stated and providers named by category per the same check, with no full subprocessor list (10 of 20). The checklist (https://www.anchorterminal.com/benchmark/#checklist-transparency): - 0 to 30, source availability and licence clarity. 30 for open source under an OSI licence, 15 for closed with clear terms, 0 for unclear terms. - 0 to 30, data handling and retention statements that agree with each other (privacy policy, DPA, retention periods, subprocessors). - 0 to 20, a deprecation policy or notices with dates. - 0 to 20, telemetry disclosed with an opt-out (local software), or subprocessors and data locations disclosed (hosted). The other half of Transparency and trust is the provenance score, computed from checked facts (below). The category score is the mean of the two. Provenance checks not met in full (half of this category, computed from checked facts): - security.txt: published but past its Expires date (5 of 10) ## What we couldn't check What we couldn't read counted as absent. Publishing it on a page a plain HTTP fetch can read (not only in a browser) lets the next check count it. - unchecked: the tools as served. `initialize` and `tools/list` answered 401 without a token; the tool count (29) and names come from the docs tables, and schemas and annotations from the CLI's bundled spec, which may differ from the server. - unchecked: Coinbase's X post of 22 October 2026, so what launched that day isn't confirmed first-hand. firstReleased is 19 June 2026, the CDP changelog's date for the remote connector, four months before the lead's date. The October docs changes (x402 tools, setup guide modified 22 October) carry no changelog entry. - unchecked: the Coinbase User Agreement, privacy policy, Advanced Trade fee schedule, the English blog URL and help.coinbase.com, which answered a Cloudflare bot check. Privacy facts are per the coinbase-cdp-agentkit check of 1 October 2026; the 11 June post was read in its Dutch-locale copy. - unchecked: which countries can use the server. The docs give no region list; stocks and futures are US products. - Process note. login.coinbase.com's robots.txt disallows every path. We fetched its two OAuth metadata documents in the same command as its robots.txt, before reading it. Nothing from them is used; the authorisation endpoints recorded here come from Coinbase's setup guide. - The AI agents overview (modified 2 October 2026) says x402 is coming soon for Coinbase MCP while the MCP page documents x402 tools. Stale docs, recorded, not deducted. - Category. No category covers trading in a brokerage account. agent-wallets (custody, spending limits, revocation, x402) is the closest, since the server acts on custodial funds within portfolios the person grants, and it sits beside coinbase-cdp-agentkit. data-providers was the alternative for the market-data tools. Judgement call. - Not coinbase-cdp-agentkit. AgentKit and Agentic Wallet give an agent its own wallet; this server acts on a person's existing Coinbase account. - Our own products. It doesn't do LocalGhost's job or letme's. It trades in one Coinbase account and pays only Coinbase's curated x402 catalogue; it doesn't route tasks between tools or rate them. Graded the normal way. ## Weaknesses - Only clients Coinbase has allowlisted can connect; other agents need the local CLI and a CDP API key - `tools/list` needs a signed-in account, so tool schemas and annotations can't be read before connecting - No numeric rate limit for the remote server, no SLA and no status component for it - No per-order size limit or server-side spending budget is documented beyond portfolio scoping and the 5 USDC x402 cap - Not in the official MCP registry, and the companion CLI is at 0.0.10 with no public repository ## What costs an agent a turn today The notes we give agents before they call it. Each one is a workaround an agent shouldn't need. - Call `coinbase_portfolios_list` first and name the portfolio on every order; an order without one falls back to the default portfolio and fails if it isn't granted - Send a fresh `client_order_id` with each order and reuse it on retry, so a dropped connection doesn't place a second order - Preview with `coinbase_orders_preview` and confirm with the person before `coinbase_orders_create`; equity previews may be unavailable - Use `ETH-USDC`-style pairs when buying with a USDC balance, and `base_size` in contracts for futures - Agree a spending budget before `coinbase_x402_fetch`, set `max_amount`, and keep the `idempotency_key` for any retry; settled payments can't be reversed ## What the review panel asked for - server-side order limits - a least-privilege setup guide - Publish the fee schedule without a login - Add a server-side spending budget - Open the allowlist - Publish tool schemas - changelog entry per tool change - published server version ## When it's done Send what changed and where it's published as a dispute (https://www.anchorterminal.com/builders/#disputes, or `POST https://www.anchorterminal.com/api/v1/contact` with `"kind": "dispute"`). Disputes are answered in public, and the listing is checked again by the same checklist. Paying for an audit or a listing claim changes nothing here.