Jev by TypeSafe AI

# Fix list: Jev

From Anchor Terminal's listing at https://www.anchorterminal.com/tools/typesafe-jev, the October 2026 research run, assessed 1 October 2026. Grade B, 62.2 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 Jev: 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: No x402, MPP or L402 (0). Per-token price published without a login, $0.042 per million input tokens with output free (20). Early access with a waitlist, and we found no free tier or free credits (0). A person signs in at the console and waits for access before creating a 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. Security & auth, 53 out of 100, up to 8.2 more on the total

Why it scored 53: Read for a model API. A bearer API key from console.typesafe.ai/keys. We found no scopes, expiry or rotation in the docs and didn't check revocation. The Python SDK validates the key early and keeps it out of logged errors (18 of 30). The models page, the privacy policy (19 November 2025) and the customer agreement (23 September 2026) all say customer data isn't used for training. No retention period is stated anywhere we read, and zero retention is for enterprise customers through sales (13 of 20). Jev's job is judging content that often comes from users. The known-limitations page says an injected instruction or a misleading framing can move the answer, and the agent skill says typed output guarantees the interface, not truth. That's a written warning with no mitigation (10 of 15). Each response reports input and output tokens and carries a request ID. We found no usage or audit logs documented (6 of 15). No security.txt (404) and no bug bounty or disclosure policy found. The trust centre at trust.typesafe.ai renders only with JavaScript, so certifications are unchecked. The DPA commits to breach notice within 72 hours (6 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.

## 3. Reliability, 60 out of 100, up to 8 more on the total

Why it scored 60: Status page at status.typesafe.ai on Better Stack, with components for the API and the console and an incident list (20). The page reaches back about 30 days, which covers Jev's whole public life since 15 September. It shows three API incidents, elevated latency on 21 September (down 12 minutes), degraded traffic on 23 September (11 minutes) and degraded traffic with elevated errors on 28 September (duration not shown), plus a console outage and a 30-minute maintenance on 29 September, and gives the API 99.826 per cent over its window. No major outage, but three incidents in two weeks and less than 90 days of record (15 of 30). Limits published, 40 requests and 100,000 tokens a second, with a warning that they adjust dynamically and can change without notice (12 of 15). The API page says to retry 429 and 529 with exponential backoff, and the SDKs retry 408, 429 and 5xx with jittered backoff and honour `Retry-After` and `retry-after-ms`. A call has no side effects, so a retry is safe, but the docs don't say whether the server sends `Retry-After` (13 of 15). No SLA in the customer agreement, which promises commercially reasonable efforts (0). Early access with a waitlist (0).

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. Transparency & trust, 58 out of 100, up to 3.7 more on the total

Made of editorial 41, provenance 75.

Why it scored 58: Closed model under a customer agreement, terms of use and an acceptable-use policy, with MIT SDKs (15). The privacy policy, the DPA (24 April 2026) and the customer agreement agree on no training. None gives a retention period. The policy keeps data as long as reasonably necessary, and the agreement lets TypeSafe delete customer data after termination at its discretion. The privacy policy names Google Analytics (14 of 30). Version pinning is documented, but there's no deprecation policy, and the agreement's notice promise is commercially reasonable efforts (6 of 20). Services are hosted in the United States, per the privacy policy. The DPA points to a subprocessor list at trust.typesafe.ai/subprocessors that renders only with JavaScript, so we couldn't read it (6 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: typesafe.ai, no registry record we could read (0 of 15)
- security.txt: not found (0 of 10)

## 5. Maintenance & community, 64 out of 100, up to 3.2 more on the total

Why it scored 64: Read for a closed model. Python SDK 0.7.2 on 26 September, and Jev went public on 15 September (30). Model aliases and pinning advice exist, but the customer agreement promises only commercially reasonable efforts to give advance notice of API changes, and the site terms allow changes without notice (5 of 12). One model version in two and a half weeks of public life, too short to show churn either way (4 of 8). Public SDK changelogs that call out breaking changes (0.6.0 and 0.7.0). Issues are welcome but pull requests aren't accepted, and issue replies are unchecked because GitHub refused our reader (7 of 15). Current SDKs in Python and TypeScript, though TypeScript is at 0.6.0 against Python's 0.7.2 (13 of 15). The public repositories are mirrors published by a bot. Actions are pinned by commit and the publish job builds and checks the package, but no public workflow runs the test suites (5 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.

## 6. Agent ergonomics, 84 out of 100, up to 2.6 more on the total

Why it scored 84: Read as an API an agent calls for a decision, since Jev has no tool calling or text generation to grade. Answers are a few numbers per question. The SDK's recorded live call answered three questions with 55 output tokens from 448 input tokens. The cost is the state, capped at 32,000 tokens with the longest question inside a 64,000-token request, and we found no prompt caching or batch endpoint (18 of 25). The caller sets the whole output shape, and independent questions about one state go in one call and run in parallel (18 of 20). The SDKs raise a typed exception per status (400, 401, 403, 404, 422, 429 and 5xx) carrying the `x-typesafe-request-id` header, and format 422 details with the failing field's path. The API page doesn't document the error body (15 of 20). A decision has no side effects, and the SDKs retry 408, 429 and 5xx twice by default with backoff and `Retry-After` (18 of 20). Python and TypeScript SDKs, `jev-latest` as the default model, and `instructions` optional on every question (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. Schema & documentation, 87 out of 100, up to 2.1 more on the total

Why it scored 87: OpenAPI 3.1 at api.typesafe.ai/openapi.json (TypeSafe v0.2.0), with `POST /v1/systemone`, `GET /v1/models` and a bearer scheme, and the Python SDK's wire models are generated from it (25). llms.txt indexes more than 100 pages, each with a Markdown twin at `.md` (10). The docs and the agent skill say when to use Jev (routing, ranking, extraction, verification), and a known-limitations page for Jev 1.13 says when not to, naming arithmetic, counting, date comparison, double negatives, large irrelevant state, adversarial content and text generation (18 of 20). Three typed question kinds behind a `type` enum, up to 255 labels per choice and 2 to 10 levels per score. The OpenAPI sets only a minimum of one question, so the 255 and 10 limits live in prose, and state is free-form by design (12 of 15). curl and SDK examples and 18 cookbooks. Errors 401, 422, 429 and 529 are listed and the SDKs also map 400, 403, 404 and 5xx, but the error body isn't documented (12 of 15). Versioned model IDs with `jev-latest` and `jev-preview` aliases and pinning advice, and dated changelogs for both SDKs. No changelog for the API or the model (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.

## 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: issue and pull-request response times on the SDK repositories, since GitHub refused our reader
- unchecked: the subprocessor list and any certifications on trust.typesafe.ai, which renders only with JavaScript
- Whether early access needs a card, and whether approval comes with free credits
- Whether keys can be scoped, rotated or set to expire
- Whether the API sends Retry-After on 429 and 529, and whether failed requests are charged
- The registration date of typesafe.ai, since RDAP refused our reader
- Cloudflare's launch post gives Jev a 32k context. TypeSafe's models page says 64,000 tokens a request with 32,000 for state plus the longest question, and we used TypeSafe's figures

## Weaknesses

- Early access behind a waitlist, with no free tier or free credits found
- No SLA, and the customer agreement promises only commercially reasonable efforts to give notice of API changes
- Published limits of 40 requests a second can change without notice
- No security.txt and no stated retention period, and the trust centre and subprocessor list render only with JavaScript
- Closed and text only, with 32,000 tokens for state plus the longest question

## 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.

- Put every independent question about one state into a single call. They run in parallel and the state is billed once
- Pin `jev-1.13.0` instead of `jev-latest` once you've tuned confidence thresholds
- Back off exponentially on 429 and 529. The limits move with demand
- Keep state to what the decision needs. Accuracy falls as unrelated content grows, and state plus the longest question must fit in 32,000 tokens
- Treat an answer about user-supplied text as a judgement that hostile text can steer, and cap what one answer can trigger

## What the review panel asked for

- API and model changelog
- fixed notice period
- Add a free trial tier
- State failed-call billing

## 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.