Chroma API + MCP by Chroma

# Fix list: Chroma API + MCP

From Anchor Terminal's listing at https://www.anchorterminal.com/tools/chroma, the October 2026 research run, assessed 1 October 2026. Grade E, 45.4 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 Chroma API + MCP: 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, 38 out of 100, up to 10.9 more on the total

Why it scored 38: Chroma Cloud keys go in `x-chroma-token` and are tied to a tenant and database per the 30 September check, with no other scopes found. A local or self-hosted server runs with no auth unless configured (20). No read-only key, and the MCP server has no read-only mode, though a key limited to one database narrows the reach (5). Stored documents come back as written, including pages pulled in by Sync from the web, and we found no prompt-injection guidance (5). No audit log found (0). The Team plan claims SOC 2. No security.txt or bug bounty found, the MCP repo's SECURITY.md still has a "[your-email]" placeholder, and the critical advisory below has sat with no patched release since May (8).

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, 30 out of 100, up to 8.8 more on the total

Why it scored 30: No x402, MPP or L402 (0). Per-unit prices published without login, $2.50 per GiB written, $0.33 per GiB a month stored, $0.0075 per TiB scanned and $0.09 per GiB returned (20). $5 of credit for new users, but neither the pricing page nor the pricing docs say whether a card is needed (10 of 20). A person signs up in the browser to get a key. The in-process library needs no account, but the rubric scores the hosted option (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. Reliability, 60 out of 100, up to 8 more on the total

Why it scored 60: Instatus page at status.trychroma.com with four components (API, Dashboard, Package Search MCP, hosted embedding) and monthly history (20). One incident in the last 90 days, degraded indexing for some customers on 18 August 2026 for 2 hours 38 minutes, so minor only (20). Chroma Cloud quotas are published with numbers, but as concurrency and size caps (10 concurrent reads and 10 writes per collection, 300 records a write, 300 results a query), not as requests a second (10 of 15). No 429, Retry-After or backoff guidance found in the docs (0). Enterprise lists "SLAs" with no figure, and nothing for Team (0). Chroma Cloud isn't marked 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.

## 4. Maintenance & community, 35 out of 100, up to 5.7 more on the total

Why it scored 35: Newest release is the JavaScript client 3.5.0 on 30 June 2026, 93 days before this check. The server and Python package last shipped as 1.5.9 on 5 May 2026 (10). No tagged or published release since 3 July (0). The main branch is busy, with 156 commits since 1 July, but the open issue for CVE-2026-45829 has community requests for a patch release and no maintainer reply we could find (10). Python and JavaScript clients are official, but the Python package is five months old and `chroma-mcp` last shipped as 0.2.6 on 14 August 2025, pinned to chromadb 1.0.16, and isn't in the official MCP registry (10). 23 CI workflows including tests and a Python vulnerability scan, but the released package still carries a critical advisory (5).

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, 71 out of 100, up to 4.7 more on the total

Why it scored 71: OpenAPI at api.trychroma.com/openapi.json per the 30 September check (25). llms.txt indexes 162 pages, each served as Markdown (10). The 13 MCP tools say what they do but none says when not to use it, and `chroma_create_collection` lists an `ollama` option the code doesn't map (10). Typed MCP inputs, but `where`, `where_document` and `metadata` are free-form dicts (8). Filter examples sit inside the query tool's description and the docs have a troubleshooting page, but we found no error catalogue (8). `/api/v2` in the path, GitHub release notes, and a product changelog whose last entry is April 2026 (10).

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. Agent ergonomics, 73 out of 100, up to 4.4 more on the total

Why it scored 73: 13 MCP tools, in the 11 to 30 band (15). The API and the query tool take `include` to drop embeddings or documents from the response (3 back). `limit` and `offset` on list and get, `n_results`, metadata and document filters, and pagination and field selection in the Search API (20). MCP errors come back as "Failed to ... " plus the raw exception text, and we found no documented error codes (10). `upsert` makes writes safe to repeat, but the MCP tools carry no readOnlyHint or destructiveHint, including `chroma_delete_collection` (10). Only the collection name and query text are required. Official Python and JavaScript clients (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.

## 7. Transparency & trust, 75 out of 100, up to 2.2 more on the total

Made of editorial 68, provenance 82.

Why it scored 75: Apache-2.0 (30). The privacy policy names Chroma Inc., links a subprocessor list at trychroma.com/subprocessors and a DPA, but it was last updated 2 October 2024, says the service is hosted in the United States although the changelog announced an EU region in April 2026, and states no retention period beyond "no longer necessary" (18). No deprecation policy or dated notices found, only a migration guide (5). Subprocessors listed. In the current source the PostHog product-telemetry client is a no-op, and OpenTelemetry stays with the operator (15).

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: trychroma.com, registered 2022-03-23 (4 years) (7 of 15)
- 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.

- CVE-2026-45829 (GHSA-f4j7-r4q5-qw2c), published 18 May 2026, critical with CVSS 9.3. An unauthenticated request to the collections endpoint can set `trust_remote_code` and run code on a chromadb server from 1.0.0 to 1.5.9. The fix merged to main on 7 July 2026, but no release carries it, and the advisory still lists no patched version (https://github.com/advisories/GHSA-f4j7-r4q5-qw2c, https://pypi.org/project/chromadb/#history). We found nothing saying whether Chroma Cloud was exposed, so we deduct for the unpatched self-hosted path only.

## 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 Chroma Cloud's $5 starter credit needs a card
- Whether Chroma Cloud was ever exposed to CVE-2026-45829, and when a patched chromadb release will ship
- Whether the July tenant-scope fix in the cloud frontend closed a reachable cross-tenant path or only added a second check
- Whether Chroma Cloud returns 429 with Retry-After when the concurrency caps are hit

## Weaknesses

- CVE-2026-45829, CVSS 9.3, has no patched release more than four months after the advisory
- No tagged or published release since 30 June 2026, and the server last shipped on 5 May
- `chroma-mcp` last released on 14 August 2025, with 13 tools and no read-only mode or tool annotations
- Cloud caps of 300 results a query, 300 records a write and 10 concurrent reads per collection, with no 429 or retry guidance
- Privacy policy dates from October 2024 and still says the service is hosted only in the United States

## 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.

- Don't expose a self-hosted chromadb 1.5.9 server, which has no auth by default and an unpatched code execution flaw
- Batch writes in groups of 300 or fewer on Chroma Cloud
- Use the Search API for hybrid ranking. The older `query()` only does dense search plus filters
- Pass `include` without embeddings, since returned GiB are billed
- Don't pass `ollama` to `chroma_create_collection`. The description lists it but the server can't map it

## What the review panel asked for

- a CVE-2026-45829 patch release
- a release schedule
- State failed-call billing
- Say if credit needs card

## 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.