Zep by Zep

# Fix list: Zep

From Anchor Terminal's listing at https://www.anchorterminal.com/tools/zep, the October 2026 research run, assessed 1 October 2026. Grade B, 69.6 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 Zep: 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, 30 out of 100, up to 8.8 more on the total

Why it scored 30: No x402, MPP or L402 (0). Credits are priced per unit and public, 1 credit per 350 bytes of episode, overage $25 per 10,000 on Flex and $75 per 40,000 on Flex Plus (20). Free has 10,000 credits a month, but the pricing page doesn't say whether a card is needed (10 of 20). A person signs up in the browser and creates a project key, and the MCP server needs a human identity-provider sign-in (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, 65 out of 100, up to 7 more on the total

Why it scored 65: Atlassian Statuspage at status.getzep.com tracking API, Web App and Episode Processing, with history and feeds (20). The last 90 days hold four incidents. Two impaired core functions for over an hour, retrieval and episode processing on 27 August (about 2.5 hours) and errors and latency during the AWS us-west-2 outage on 3 September (about 2 hours). Two were search latency on 10 and 15 September (2.5 and 3.5 hours). We count two majors as 5, between the rubric's 10 for one and 0 for several. Limits published per plan, 600 requests a minute on Flex and 1,000 on Flex Plus, with a 5 RPM penalty mode on Free (15). 429 with Retry-After, X-RateLimit headers and backoff-with-jitter guidance, but no idempotency or safe-retry advice for writes (10 of 15). Enterprise names "guaranteed rate limits with SLA" with no published figure (5 of 10). Generally available (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, 70 out of 100, up to 4.9 more on the total

Why it scored 70: The Memory MCP has 12 tools, 10 read and 2 write (15), and an administrator can switch a connection to read-only (5 added back). Search takes a scope (edges, nodes, episodes and more) and a limit, and add-messages can return the context block in the same call. We didn't confirm cursor pagination on list calls (15 of 20). 429s come back as typed SDK errors with headers, and endpoints document 400 and 500 (14 of 20). No idempotency keys found, and MCP tools are described as read or write rather than carrying confirmed annotations (6 of 20). SDKs in Python, TypeScript and Go, and graph adds need only data and type (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. Transparency & trust, 61 out of 100, up to 3.4 more on the total

Made of editorial 40, provenance 82.

Why it scored 61: Closed service under clear terms, with Apache-2.0 examples and SDKs (15). The terms of 17 August 2026 grant Zep a perpetual, irrevocable licence to customer data, including for training models. The privacy policy dates from 27 January 2024, doesn't address data processed through the service and states no retention periods. The two don't conflict, but together they say little (10 of 30). The terms promise 30 days' notice before changes that materially reduce core function, and the February 2026 deprecation wave has a migration page, but no standing policy (12 of 20). No subprocessor list or data-location statement found. The trust centre didn't render for us, and the only location evidence is a status note naming AWS us-west-2 (3 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: getzep.com, registered 2023-05-08 (3 years) (7 of 15)
- security.txt: not found (0 of 10)

## 5. Security & auth, 84 out of 100, up to 2.8 more on the total

Why it scored 84: Project API keys can carry ABAC policies that limit which actions and context each agent reaches, and the MCP server uses OAuth 2.1 with PKCE through the customer's identity provider. We found no key expiry or rotation documentation (27 of 30). MCP connections can be set read-only by an administrator and ABAC applies least privilege, but we found no confirmation step for deletes (15 of 20). A memory-security guide treats every context block as untrusted data and covers write control, provenance and keeping instructions out of memory (15). Audit logs for member, key, project and data operations, and API logs kept 1 day on Flex, 7 on Flex Plus and a year on Enterprise (15). SOC 2 Type II, HIPAA BAA on Enterprise and BYOK, but no security.txt, disclosure policy or bug bounty found (12 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.

## 6. Schema & documentation, 89 out of 100, up to 1.8 more on the total

Why it scored 89: OpenAPI JSON and YAML for v3 linked from llms.txt (25). llms.txt with per-section indexes, plus a docs MCP server (10). SDK reference pages say what each call does, and the MCP tools are labelled read or write (14 of 20). Typed inputs with enums (text, json, message, fact_triple) and stated limits, such as document_id 1 to 100 characters and at most 10 metadata keys (13 of 15). Examples in three SDK languages, and 400, 500 and 429 responses documented, though not every code on every page (12 of 15). v2 and v3 docs and a changelog of 54 dated entries (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.

## 7. Maintenance & community, 80 out of 100, up to 1.8 more on the total

Why it scored 80: npm @getzep/zep-cloud 3.30.0 on 2026-09-24 and PyPI zep-cloud 3.29.0 on 2026-09-23 (30). Six stable PyPI releases since 10 July and 20 changelog entries in September alone (20). Closed service, scored on the 15-point scale. The changelog is public and frequent, and we didn't test a support channel (10 of 15). Current official SDKs for Python, TypeScript and Go (15). A 4.0 alpha series started in August, and we didn't check CI on the SDK repos (5 of 10). Zep's own servers aren't in the official MCP registry.

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.

- Whether API keys can be rotated or set to expire isn't documented in the pages we read.
- trust.getzep.com didn't render, so the subprocessor list and data regions are unverified.
- Whether the Memory MCP tools carry readOnlyHint or destructiveHint annotations isn't stated.
- The listing said OpenAPI wasn't public. It is, at help.getzep.com/v3/openapi.json, and the patch sets it.
- The listing said Free needs no card. The pricing page we read on 2026-10-01 doesn't say either way, so we scored it as not stated and patched the text.

## Weaknesses

- The terms of 17 August 2026 grant a perpetual, irrevocable licence to train models on customer data
- First paid plan is $125 a month, and credits grow with episode size
- The MCP server needs OAuth through a company identity provider, so a headless agent can't use it
- Four status incidents between 27 August and 15 September 2026
- No security.txt, no subprocessor list and no entry in the MCP registry

## 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.

- Create the user and a thread before adding messages, as the quickstart does
- Set `return_context` when adding messages to skip a second round trip
- Keep episodes short. Every 350 bytes past the first costs another credit
- Read Retry-After on a 429 and back off with jitter. Free accounts over quota drop to 5 requests a minute
- Expect a delay after a graph add (202 Accepted) before new facts show up in search

## What the review panel asked for

- Add tool annotations
- a training opt-out
- key expiry and rotation

## 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.