# Fix list: ObsessionDB From Anchor Terminal's listing at https://www.anchorterminal.com/tools/obsessiondb, the October 2026 research run, assessed 21 October 2026. Grade E, 41.1 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 ObsessionDB: 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, 20 out of 100, up to 16 more on the total Why it scored 20: Graded as a hosted service. No status page was found. None is linked from the site, status.obsessiondb.com did not answer through our proxy, and the home page sets its contractual SLAs against 'a status page' (0). No readable incident history (5). No request rate limit found for the MCP server or the ClickHouse endpoints. The 1,000-row or 10 MiB `run_sql` cap and the one-insert-a-second guidance are result and usage guidance, read under ergonomics (0). No 429 or backoff guidance. Insert deduplication is handled by the coordination layer and Jobs runs large SQL as retryable tasks (5 of 15). The terms (10 June 2026) give no contractual uptime SLA on self-service plans and the enterprise SLAs are not published (0). The MCP server and the managed service carry no beta or preview label, though Datashares are in beta and self-service tier choice is 'releasing shortly' (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. Schema & documentation, 43 out of 100, up to 9.3 more on the total Why it scored 43: No public contract for the MCP tools. Their input schemas come back only from `tools/list` with a signed-in token, which we didn't obtain, and the ClickHouse HTTP interface has no OpenAPI file. The OAuth metadata documents are machine-readable but don't describe tools (0). llms.txt and a Markdown copy of each docs page (10). The docs give one line per tool, say to call `list_services` first, explain slug clashes and when to push work into SQL. ObsessionDB's post of 18 June 2026 quotes one tool calling itself 'the entry point for service-wide error investigation' (10 of 20). Inputs are a service slug and a free-text SQL string, with four documented scopes. No enums or constraints could be read (5 of 15). Setup examples for eleven clients, and documented `truncated: true` and `available: false` results. ClickHouse errors come back in a field per the 18 June post, with no error catalogue. The stdio bridge is called 'Anthropic's `mcp-remote` proxy', and npm names Glen Maddern as its author (8 of 15). A dated public changelog with 19 entries that tracks the ClickHouse release (26.8 LTS), with no version for the MCP server (10 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. ## 3. Security & auth, 51 out of 100, up to 8.6 more on the total Why it scored 51: OAuth 2.1 with dynamic client registration, S256 PKCE, refresh tokens, header-only bearer tokens and four scopes, revocable from the console. The authorisation server metadata was read on 21 October 2026. Less 4 because `workbench:query` runs any SQL as the person's account across every organisation they belong to, with no per-service scope (26 of 30). The MCP docs describe no read-only mode or confirmation step for `run_sql`. ClickHouse roles down to table and column, a global read-only role since May 2026 and read-only Datashares give least-privilege options to an operator who sets them up (8 of 20). `run_sql` returns stored rows to the model as they are, and no prompt-injection guidance was found. The agent context button masks passwords and keys in captured SQL (2 of 15). ClickHouse's `system.query_log` records each query and the console keeps a one-week monitoring archive. Whether MCP calls are attributed to a client was not found (10 of 15). No SOC 2 or ISO 27001 (stated on the security page), no security.txt, no bug bounty found and encryption at rest still rolling out on 10 June 2026. The security page names an address for vulnerability reports (5 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. ## 4. Agent ergonomics, 53 out of 100, up to 7.6 more on the total Why it scored 53: Twelve tools with short descriptions, all but `run_sql` read-only by name. Whether a person can approve fewer scopes at consent was not found (15). `run_sql` stops at 1,000 rows or 10 MiB and flags truncation, `monitoring_list_query_executions` filters to failed runs, and pattern tools take a time window. No paging was documented (12 of 20). Errors return in a field rather than failing the call, monitoring tools answer `available: false` where there is no archive, and a slug clash returns a rename request. ClickHouse error codes are documented upstream (12 of 20). No `readOnlyHint` or `destructiveHint` could be read from the docs, and no idempotency guidance for `run_sql` (4 of 20). Two required inputs for most calls. ObsessionDB ships no SDK of its own; ClickHouse's own drivers in Go, Python, JavaScript, Rust, .NET and Java are documented with per-driver setup pages (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. Payments & pricing, 40 out of 100, up to 7.5 more on the total Why it scored 40: No x402, MPP or L402 (0). Compute at $20 per GB of memory per node per month and storage at $25 per TB per month are public on the pricing page without a login, which calls them indicative (20). A free developer tier of up to 4 vCPU and 8 GB with 'No credit card required' (20). A person signs up in a browser and approves OAuth consent before an agent can connect (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. ## 6. Maintenance & community, 28 out of 100, up to 6.3 more on the total Why it scored 28: Graded as a closed service. The last dated changelog entry is ClickHouse 26.8 LTS on 31 August 2026, 51 days before the check (20). Two dated entries in the 90 days to 21 October 2026, 21 and 31 August, short of three (0). A public changelog, in-console live chat since June 2026, a shared Slack channel for customers and email support. Replies were not sampled (8 of 15). Not in the official MCP registry and no SDKs of its own (0). Nothing to install for the service. chkit, its open-source CLI, had 93 commits in the 90 days and CI workflows on its default branch (0 of 10 for the service, as there's no package to check). 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, 58 out of 100, up to 3.7 more on the total Made of editorial 50, provenance 65. Why it scored 58: Closed service with published terms under German law. The engine is open-source ClickHouse and the storage layer is proprietary (15). A privacy policy covering the website and the service, a subprocessor list with locations and transfer bases, a seven-day deletion target on offboarding and 14-day metadata backups. The DPA is on request only. The home and pricing pages say fully managed on AWS, Google Cloud and Azure in every public region, while the subprocessor list of 10 June 2026 names Hetzner and Backblaze and none of those clouds (16 of 30). The terms promise reasonable notice before a material reduction of core functions on a paid service. No deprecation notices with dates were found (5 of 20). Eleven subprocessors with purpose and location and 30 days' notice of changes, less for the missing cloud providers (14 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: obsessiondb.com, registered 2026-01-03 (under a year) (0 of 15) - Status page: not found (0 of 10) - security.txt: not found (0 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 MCP tool input schemas and annotations, which `tools/list` returns only to a signed-in client. - The 21 October 2026 launch claims (100 times faster and 20 times cheaper than Snowflake and Databricks, trillions of rows a day) were relayed to us from X and are not on obsessiondb.com. No independent benchmark shows them. ClickBench's repository (commit of 20 October 2026) has no ObsessionDB entry, and the raw results promised on GitHub in the 18 February 2026 post were not found. - The home and pricing pages say fully managed on AWS, Google Cloud and Azure in every public region; the subprocessor list of 10 June 2026 names only Hetzner and Backblaze for compute and storage. Which providers hold data for a cluster on those clouds was not established. - No status page was found; status.obsessiondb.com did not answer through our proxy, so a page on another host can't be ruled out. - Whether a person can grant fewer than all four scopes at consent, and whether MCP calls are attributed in `system.query_log`, was not found. ## Weaknesses - No public status page or incident history was found, and the terms give no uptime SLA on self-service plans - No SOC 2 or ISO 27001, no security.txt, and encryption at rest was still rolling out per the security page of 10 June 2026 - The `workbench:query` scope runs any SQL as the person's account, with no read-only MCP mode documented - Only two dated changelog entries in the 90 days to 21 October 2026, the last on 31 August - Speed and cost claims come from ObsessionDB's own runs. It has no entry in ClickBench's results 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 `list_services` first and pass `serviceSlug` unchanged from it to `run_sql` and the monitoring tools - Push aggregation into SQL with `GROUP BY` or `ORDER BY ... LIMIT`. Results stop at 1,000 rows or 10 MiB and set `truncated: true` - Treat `available: false` from a monitoring tool as no monitoring archive for that service, not as an error - Batch inserts to about one `INSERT` a second per table, because every insert writes durably to object storage - Expect a browser sign-in on first connect. A person must approve the OAuth consent at console.obsessiondb.com ## What the review panel asked for - firm rate card per region - publish the tool schemas - version the MCP server - publish rate limits and 429 behaviour - a public status page - a read-only MCP scope - per-service scopes ## 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.