Graphiti by Zep

# Fix list: Graphiti

From Anchor Terminal's listing at https://www.anchorterminal.com/tools/graphiti, the October 2026 research run, assessed 1 October 2026. Grade D, 53.3 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 Graphiti: 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, 23 out of 100, up to 13.5 more on the total

Why it scored 23: Scored for a framework you run yourself. The LLM and database credentials come from environment variables, and nothing travels in a URL, but the streamable HTTP endpoint on port 8000 has no authentication in the server code, so anything that can reach the port can call clear_graph (15 of 30). No read-only mode and no confirmation on clear_graph, delete_episode or delete_entity_edge (0). Facts and episodes come back from whatever was ingested, and we found no injection guidance for Graphiti (0). No audit log of tool calls found (0). SECURITY.md is in the repo. We didn't find published advisories either way, and there's no bug bounty (8 of 20). Telemetry is on by default, anonymous and documented, with GRAPHITI_TELEMETRY_ENABLED=false to turn it off.

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. Reliability, 50 out of 100, up to 10 more on the total

Why it scored 50: Scored on the local-package checklist. graphiti-core installs from PyPI with Python 3.10 to below 4 stated (20). GitHub Actions run lint, unit tests and MyPy, but we didn't confirm the default branch is green (15 of 25). 255 open issues and 176 open pull requests, with bug reports from June and July 2026 still open, among them a broken MCP Docker build (#1624) and edge-search reranking dropping candidates (#1642). We didn't confirm maintainer reply times (10 of 25). Version numbers follow 0.x semver, but the GitHub releases page stops at v0.29.2 while PyPI has 0.29.3, 0.30.1 and 0.30.2, so the newest changes have no release notes there (5 of 15). Still below 1.0 (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.

## 3. Agent ergonomics, 51 out of 100, up to 8 more on the total

Why it scored 51: 13 MCP tools (15), with no toolsets or read-only subset. Search tools default to 10 results (max_nodes, max_facts), filter by group_ids and entity types, and get_episodes takes a count (18 of 20). Error shapes aren't documented and we didn't test them (8 of 20). None of the 11 tools defined in graphiti_mcp_server.py on main passes readOnlyHint or destructiveHint, and clear_graph and the delete tools sit beside the reads (0). One Docker Compose command starts the server with FalkorDB and OpenAI as defaults, but the library is Python only (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.

## 4. Payments & pricing, 60 out of 100, up to 5 more on the total

Why it scored 60: Free Apache-2.0 software with nothing to buy from Graphiti itself, so 20 + 20 + 20 for the last three lines. No payment protocol (0). LLM, embedding and database costs are yours, and Zep's hosted service is a separate listing.

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.

## 5. Schema & documentation, 71 out of 100, up to 4.7 more on the total

Why it scored 71: The MCP server's 13 tools are typed Python functions, so FastMCP generates JSON Schema for every input (25). Graphiti's docs are in help.getzep.com/llms.txt (10). Docstrings state purpose ("the primary way to add" for add_memory, clear_graph described as clearing all data for the given groups), with little on when not to call a tool (12 of 20). Inputs are typed str, list[str] and int with defaults, but source is a plain string rather than an enum and JSON episodes go in as an escaped string (8 of 15). README and docs examples. We didn't find documented error responses (8 of 15). Version numbers and GitHub release notes up to v0.29.2, none on GitHub for the three later PyPI releases (8 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. Maintenance & community, 65 out of 100, up to 3.1 more on the total

Why it scored 65: graphiti-core 0.30.2 on PyPI on 2026-09-08 (30). Three stable releases since 3 July, 0.29.3 on 27 July, 0.30.1 on 1 September and 0.30.2 on 8 September (20). 255 open issues and 176 open pull requests, with reply times we didn't verify (10 of 25). The official MCP registry lists no Graphiti server from Zep (0). CI exists, Kuzu support is deprecated, and the MCP Dockerfile build broke against falkordb:latest in July (#1624) (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.

## 7. Transparency & trust, 71 out of 100, up to 2.5 more on the total

Made of editorial 78, provenance 63.

Why it scored 71: Apache-2.0 (30). Runs on your machine, and the telemetry statement says it never sends API keys, queries, episodes, nodes, edges, IP addresses or file paths. Your data goes to the LLM provider you configure (20 of 30). Kuzu is marked deprecated in the README with no removal date (8 of 20). Telemetry disclosed in the README with an environment-variable opt-out and automatic opt-out under pytest (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)
- 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.

- We didn't confirm whether CI on main is passing.
- The source on main defines 11 tools with @mcp.tool while the README lists 13 (clear_graph and get_status are the difference). We kept 13.
- We didn't confirm maintainer response times on the open issues.
- Whether the Docker Compose file binds the MCP port to localhost only isn't stated in the README.

## Weaknesses

- Every add runs LLM extraction, so ingestion costs tokens and time
- The MCP HTTP endpoint has no authentication, and no tool carries readOnlyHint or destructiveHint
- Still 0.x, and the last three PyPI releases have no GitHub release notes
- 255 open issues, including a broken MCP Docker build reported in July 2026
- No official 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.

- Set OPENAI_API_KEY (or another provider's key) and MODEL_NAME before starting the MCP server
- Use search_memory_facts for facts and search_nodes for entities instead of reading whole episodes
- Never call clear_graph to fix one fact. Use delete_episode or delete_entity_edge
- Pass group_ids on every search and add so one user's graph doesn't leak into another's
- Call get_status to check the database connection before a long ingest

## What the review panel asked for

- Add destructiveHint to delete tools
- Document the error shapes
- auth on HTTP transport
- a read-only tool mode

## 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.