# Fix list: Karrio From Anchor Terminal's listing at https://www.anchorterminal.com/tools/karrio, the October 2026 research run, assessed 8 October 2026. Grade D, 48.7 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 Karrio: 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. Security & auth, 44 out of 100, up to 9.8 more on the total Why it scored 44: Private API keys with a label, bound to test or live mode, sent as the Basic username or in the `Authorization` header. The docs say a private key can perform any request without restriction. Short-lived resource tokens for document links can travel in a `?token=` query string, which we didn't count against the score because they expire in five minutes by default and open one resource (20 of 30). Test-mode keys keep building away from live carriers. Team permissions are in the paid `ee/` code, and the MCP guide asks the agent to confirm with the user before `create_shipment` (7 of 20). Responses carry carrier tracking text and addresses, and no injection guidance was found (3 of 15). API logs, tracing records and events are stored and readable in the open-source edition, and audit logging is listed as an enterprise feature (10 of 15). `SECURITY.md` gives an email address only. No security.txt, bug bounty or certification was found, and GitHub lists no published advisories although 2026.1.32 fixed MD5 password storage (4 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. ## 2. Payments & pricing, 40 out of 100, up to 7.5 more on the total Why it scored 40: Scored with the self-hosted rule, taking prices from the paid options beside the free edition. No x402, MPP or L402 (0). Self-hosting is free, and the platform page gives starting prices only, $499 a month for the managed Scale platform with pay-as-you-go usage and $50,000 a year for a commercial licence, with no unit price (10 of 20). The open-source edition needs no card or account (20). Install is scriptable with Docker Compose and `POST /api/token` issues a JWT for an email and password, but carrier accounts and the API key are set up by a person, so half (10 of 20). 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. Agent ergonomics, 57 out of 100, up to 7 more on the total Why it scored 57: List calls take `limit` from 1 to 100 with `offset`. No field selection was found, and a shipment response nests addresses, parcels, rates and documents. The MCP server in source has 12 compact tools but isn't published (13 of 25). Offset pagination with `count`, `next` and `previous`, and filters on carrier, status, dates, keyword and metadata (17 of 20). Errors come back as an `errors` array with a code, message and details, and carrier messages pass through. The published guide lists HTTP statuses, not codes (12 of 20). No idempotency keys were found in the server or the spec. A cancel on an already cancelled shipment returns 202, and the MCP tools in source carry `readOnlyHint` and `destructiveHint` (8 of 20). A shipment created with a `service` buys the label in one call and addresses can be nested. The Python SDK is current, while the npm client `karrio` is at 2023.1.0 (7 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. ## 4. Maintenance & community, 29 out of 100, up to 6.2 more on the total Why it scored 29: The newest release, 2026.1.32, is dated 23 June 2026, 107 days before the check, and nothing has landed on main since (10 of 30). No release between 10 July and 8 October 2026 (0 of 20). Of the 15 newest Discussions threads, 10 showed no maintainer reply, and fix pull requests opened between 23 September and 2 October 2026 each had one or two comments and were unmerged, among 83 open pull requests (8 of 25). No entry in the official MCP registry and no MCP package on npm. The PyPI packages match the last release and the npm client is at 2023.1.0 (5 of 15). CI passed on the last commit and the changelog records pinning axios away from compromised versions, while dependency pull requests from April and June 2026 remain open (6 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. ## 5. Schema & documentation, 68 out of 100, up to 5.2 more on the total Why it scored 68: `schemas/openapi.yml` is OpenAPI 3.0.3 at version 2026.1.32, with 65 paths, and the docs site renders a reference from it. GraphQL schemas sit beside it (25). No llms.txt on www.karrio.io or docs.karrio.io, both 404 (0). The spec carries about 1,200 description lines and the guides cover authentication, pagination, shipments, trackers and webhooks, with little on when not to use a call, such as the stored `/v1/shipments` against the pass-through `/v1/proxy/shipping` (12 of 20). Units, statuses and label types are enums and required fields are marked, but carrier `options` and connection `credentials` are open objects whose keys differ by carrier (10 of 15). The guides have cURL and JavaScript examples and the spec has an `ErrorResponse` schema, while the error guide is a table of HTTP statuses only (9 of 15). Versions are dated and `CHANGELOG.md` is current, but the changelog page on the website stops at 2024.12.6 (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. ## 6. Reliability, 76 out of 100, up to 4.8 more on the total Why it scored 76: Read with the local-software lines, since the graded surface is the open-source server its owner hosts. The managed platform has no status page we could find, and its hosts didn't answer on 8 October 2026. Docker images on Docker Hub and PyPI packages at 2026.1.32, with Python 3.11 or later stated for `karrio`, though `karrio-server` still declares 3.7 or later (18 of 20). A public tests workflow runs SDK type checks, SDK tests and server tests on PostgreSQL, and the last five runs on main, all on 23 June 2026, passed (25). GitHub Issues is switched off, so bug reports go to Discussions. Reports from July and August 2026 on FedEx dimensions and a GraphQL crash had no reply, and a June report that migration 0093 deleted tracker history was fixed in 2026.1.32 (10 of 25). Versions are calendar-based, not semver, and the breaking rename of the `purchased` status shipped in 2026.1.20 with a banner in the changelog (8 of 15). Releases are past 1.0 and numbered for production use (15). 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. ## 7. Transparency & trust, 56 out of 100, up to 3.9 more on the total Made of editorial 59, provenance 53. Why it scored 56: The server, SDK and modules are LGPL-3.0, an OSI licence. `ee/` is under a proprietary licence in the same repository, and the PyPI metadata for `karrio` leaves the licence field empty (26 of 30). On a self-hosted instance the data stays with the owner. The privacy policy (8 April 2025) covers the website and associated services, excludes customer data, gives no retention periods and names 12 service providers. No data processing addendum was found (14 of 30). Breaking changes are flagged in the changelog and the spec says a new dated version marks an incompatible change. No deprecation policy with notice periods was found (7 of 20). Sentry and PostHog reporting are documented and stay off unless the operator sets their keys. The Compose files pull images through a Scarf gateway and the README carries a Scarf pixel, neither described in the docs we read (12 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): - Domain age: karrio.io, registered 2022-01-20 (4 years) (7 of 15) - Endpoint on the vendor's domain: is not on karrio.io (0 of 15) - Terms of service: read, states 5 of the 7 things a reader expects, and has 1 clause that costs points (6.3 of 10) - Status page: not found (0 of 10) - security.txt: not found (0 of 10) ## Deductions Each comes off the total. A fixed and documented problem counts for less at the next check. - 12 March to 23 June 2026. Versions 2026.1.22 to 2026.1.31 applied an MD5 `PASSWORD_HASHERS` override outside the test runner, so production instances stored MD5 password hashes. Fixed in 2026.1.32 and described in the changelog, with no published advisory, so 4 of a possible 15 (https://github.com/karrioapi/karrio/blob/main/CHANGELOG.md). - 8 October 2026. The README and the MCP guide tell users to run `npx karrio-mcp` and `npx -y @karrio/mcp`, and the npm registry answers 404 for both. The examples' API host `api.karrio.io` didn't resolve the same day. 3 for a claim the listing can't meet (https://registry.npmjs.org/@karrio%2fmcp, https://www.karrio.io/docs/developing/mcp-server). ## 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: whether the managed platform is still operating. `api.karrio.io` didn't resolve and `app.karrio.io` and `platform.karrio.io` returned Vercel's paused-deployment error on 8 October 2026, which may be temporary - unchecked: the community carrier submodule, which we didn't fetch, so the connector count of 30 covers `modules/connectors` only - unchecked: reply times in Discord, and the sponsorship prices for Insiders, since polar.sh/karrioapi returned 404 - Whether releases have moved to a private or Insiders repository since June 2026 wasn't established - No address validation endpoint was found in the OpenAPI file, so `shipping.address-validation` is left out. Returns rest on the `is_return` and `return_shipment` fields in the spec, which we didn't exercise - The lead said REST API and SDKs with a cloud option. The cloud option couldn't be reached, and the listing grades the self-hosted server ## Weaknesses - No release since 2026.1.32 on 23 June 2026, 107 days before this check, and fix pull requests opened in September 2026 were still open - Neither `@karrio/mcp` nor `karrio-mcp` exists on npm, though the README and the MCP guide give `npx` commands for them - `api.karrio.io` didn't resolve and `app.karrio.io` answered 503 with a paused deployment on 8 October 2026, and the managed platform starts with a demo booking - Versions 2026.1.22 to 2026.1.31 stored passwords as MD5 hashes in production, fixed in 2026.1.32 with a changelog line and no published advisory - No idempotency keys were found, and a private API key can perform any request. Team permissions and audit logging sit in the paid `ee/` code ## 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. - Run your own instance with Docker Compose and call it at `http://localhost:5002`. Don't rely on `api.karrio.io`, which didn't resolve on 8 October 2026 - Don't run `npx karrio-mcp` or `npx -y @karrio/mcp`. Neither name is published on npm, so build the server from `packages/mcp` in the repository - Change the default `admin@example.com` and `demo` login before the instance is reachable, and upgrade to 2026.1.32 or later so passwords aren't stored as MD5 - Don't blind-retry `POST /v1/shipments` with a `service` set or `POST /v1/shipments/{id}/purchase`. Fetch the shipment and check its status first, since no idempotency key exists - Use a test-mode API key while building. Send it as the HTTP Basic username or as `Authorization: Token key_...` ## 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.