Dust

# Fix list: Dust

From Anchor Terminal's listing at https://www.anchorterminal.com/tools/dust, the October 2026 research run, assessed 8 October 2026. Grade B, 62.8 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 Dust: 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. Payments & pricing, 20 out of 100, up to 10 more on the total

Why it scored 20: Read with the hosted rubric. No x402, MPP or L402 (0). Plan prices are public per seat with their credit allowances, but no price per top-up credit was found and the Enterprise programmatic rate is by quote (10). A Free seat has 500 lifetime credits, and the pricing page does not say whether a card is needed. Programmatic usage has no free credits and draws only on the workspace pool, which Free seats cannot use, so the free tier does not cover API calls that run agents (10 of 20). A person signs in through a browser to create a workspace and an admin creates the key (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.

## 2. Reliability, 53 out of 100, up to 9.4 more on the total

Why it scored 53: Read with the hosted lines, on the REST API and the remote MCP server. status.dust.tt runs on Statuspage with 16 components, among them API, Conversations and Data Sources, and a dated history (20). In the 90 days to 8 October 2026 it lists two major incidents, degraded performance for 2 hours 3 minutes on 16 July covering Conversations, the app platform and the API, and 6 minutes of inaccessible conversations on 18 September, plus seven minor ones, mostly latency. That is one major of over an hour (10). Rate limits carry numbers for document upserts (120 a minute per workspace) and Dust app runs (10,000 a day) only, with none for conversations or search (8 of 15). The OpenAPI document lists 429 on six operations, and no Retry-After or backoff guidance was found in the docs. Upserts by document ID are safe to repeat (5 of 15). No SLA was found on the pricing or security pages, and the terms went unread (0). The REST API carries no beta label, while the MCP server is described as a first version and the client-side MCP server is a 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.

## 3. Agent ergonomics, 67 out of 100, up to 5.4 more on the total

Why it scored 67: Workspace search takes `limit` (1 to 100, default 25), and data source search takes `top_k` and `target_document_tokens`, so results can be sized. The MCP server's tool list needs a signed-in workspace and was not read (20 of 25). Cursor pagination on search, `limit` and `lastValue` on conversations, and tag, parent and timestamp filters on data source search (20). The source returns errors as a type and a message, such as `rate_limit_error` with the reason a call was blocked, but the docs give no list of error types (12 of 20). No idempotency keys in the OpenAPI document. Document and row upserts by ID can be repeated safely, and MCP tool annotations were not seen (6 of 20). Search needs only a query. One official SDK language, JavaScript, plus a CLI with a non-interactive mode that prints JSON (9 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. Security & auth, 75 out of 100, up to 4.4 more on the total

Why it scored 75: API keys go in the `Authorization` header, can be limited to chosen spaces and capped by spend, and the MCP server takes OAuth with dynamic client registration, client ID metadata documents and a required `resource` parameter, issuing tokens that work only for MCP (28 of 30). An MCP client works as the signed-in user with that user's spaces only, an admin can switch the server off and restrict redirect URIs, and the API has an endpoint to approve or reject an agent's action. No read-only key type was found (16 of 20). Search returns content from connected sources, and no prompt-injection guidance for API or MCP clients was found in the pages read (3 of 15). Audit logs record user, API key and system actors with SIEM streaming, on the Enterprise plan only (12 of 15). SOC 2 Type II and HIPAA support on the security page, a HackerOne disclosure programme with no bounty stated, CodeQL in the public CI, and no security.txt (16 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.

## 5. Schema & documentation, 75 out of 100, up to 4.1 more on the total

Why it scored 75: A public OpenAPI 3.0 document with 54 paths and 68 operations, linked from the docs and from llms.txt (25). llms.txt at docs.dust.tt with a Markdown copy of every page (10). Operation descriptions mostly restate the summary ("Search for nodes in the workspace"), six operations have none, and none says when not to use an endpoint (10 of 20). 43 enums and 277 required markers, though ranges such as the search `limit` of 1 to 100 sit in the description text and not in the schema (10 of 15). 186 examples, while most error responses are a status and two words ("Bad request") with no body schema (9 of 15). The path carries `/api/v1`, the document a version (1.0.2), and the product changelog is dated and has carried API deprecation notices. There is no changelog for the API alone (11 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. Transparency & trust, 68 out of 100, up to 2.8 more on the total

Made of editorial 71, provenance 64.

Why it scored 68: The platform's source is public under the MIT licence (30). The Platform Privacy Policy gives retention periods (the agreement's length plus five years archived, three years for marketing contacts) and says model providers keep no data and may not train on it, which agrees with the security page. It covers Dust as controller only, and the DPA and terms of service went unread (18 of 30). No written deprecation policy was found. The changelog gave dated notice on 8 April 2026 that two analytics endpoints would be removed on 1 June 2026, under two months ahead (10 of 20). The privacy policy names the model providers (OpenAI, Anthropic, Mistral, Google, Fireworks), Google Cloud and Qdrant Cloud, and hosting is in the US or the EU. The full sub-processor list on the trust centre is drawn by script and was not read (13 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: dust.tt, no registry record we could read (0 of 15)
- Terms of service: not found (0 of 10)
- Privacy policy: read, states 6 of the 8 things a reader expects (8.5 of 10)
- security.txt: not found (0 of 10)

## 7. Maintenance & community, 86 out of 100, up to 1.2 more on the total

Why it scored 86: The changelog's newest entry is dated 8 October 2026 and the repository had 112 commits that day (30). 49 changelog entries from August to 8 October 2026 (20). Issues are public on GitHub (388 open issues and pull requests by the API's count), with a support email and a community site. We could not read reply times (15 of 25). `@dust-tt/client` 1.2.9 was published on 22 September 2026 and the CLI on 1 September, in one language (12 of 15). Public CI builds and tests each part of the repository, with CodeQL. The SDK and CLI ask for Node 24.16 or later (9 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.

## 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 terms of service, any SLA and the DPA. dust.tt/terms redirects to dust-tt.notion.site, which is drawn by script and gave our reader no text. `provenance.terms` is left out for that reason.
- unchecked: the trust centre at trust.dust.com (certificates, sub-processor list), which is drawn by script.
- unchecked: the remote MCP server's tool count, input schemas and readOnlyHint or destructiveHint annotations, which need a signed-in workspace.
- unchecked: whether the official MCP registry lists Dust. The registry did not answer our request.
- unchecked: the effective date of the Platform Privacy Policy, which our reader did not find on the page.
- unchecked: advisories for Dust at NVD or GitHub, and reply times on GitHub issues.
- unchecked: the registration date of dust.tt. No RDAP service answers for .tt.
- Not found in the reviewed documentation: a price per top-up credit, numeric rate limits for conversations and search, and whether a Free workspace needs a card.
- The pricing page showed our reader prices in euros. Prices in other currencies were not seen, so `unitPrices` is empty.
- The lead named the client-side MCP server (preview) and left out the remote MCP server at `https://dust.tt/mcp`, which is the main MCP surface. The legal entity is Permutation Labs.

## Weaknesses

- Programmatic usage has no free credits. It draws only on the workspace credit pool, which Free seats cannot use
- Rate limits are published for document upserts and app runs only, and no Retry-After guidance was found in the docs
- The terms link redirects to a Notion site drawn by script, so the service terms, any SLA and the DPA went unread
- Audit logs are an Enterprise feature, and no security.txt is published (404)
- Two major incidents on status.dust.tt in the last 90 days, one of 2 hours 3 minutes on 16 July 2026 that covered the API

## 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.

- Use `https://eu.dust.tt` for a workspace in the EU region and `https://dust.tt` otherwise, for both the API and the MCP server
- Connect an MCP client by OAuth and send the `resource` parameter in authorisation and token requests. API keys do not work on the MCP server
- Expect HTTP 429 with type `rate_limit_error` when the credit pool, the programmatic monthly cap or the key's spend cap is exhausted. Retrying will not help
- Call `GET /api/v1/w/{wId}/search` with `limit` (1 to 100, default 25) and `cursor`, or a data source view's search with `top_k`
- Keep document upserts under 120 a minute per workspace

## 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.