Head to head · Agent memory · October 2026 research run

Agent Platform Memory Bank vs Zep

Zep scores 71 (BB) on agent readiness against Agent Platform Memory Bank's 65.6 (B), and leads in 6 of 7 scored categories. Agent Platform Memory Bank leads on transparency & trust. Both do agent memory.

Best memory layers for AI agents · All 353 memory comparisons

Which one, for what

Agent Platform Memory Bank B

Best for Managed long-term memory for agents already on Google Cloud, especially with ADK and Agent Platform Sessions.

Ahead on

  • Transparency & trust, 81 against 61

Also in its favour

  • No incidents deducted, where Zep loses 3 points for them

Watch for

Needs a Google Cloud project with billing, set up by a person in the console

Zep BB

Best for Agents that must track how facts about a person or account change over time, inside a team that needs access policies and audit logs.

Ahead on

  • Schema & documentation, 93 against 75
  • Agent ergonomics, 87 against 75
  • Security & auth, 84 against 72
  • Payments & pricing, 30 against 20
  • Maintenance & community, 80 against 62

Also in its favour

  • Agent-ready, a grade of BB or better
  • Free to start without a card

Watch for

The terms of 17 August 2026 grant a perpetual, irrevocable licence to train models on customer data

Why the edge

Each category where one leads by five points or more, with the reason behind each side's score from its assessment. The numbers are in the table below; this is what they stand for.

Transparency & trust Agent Platform Memory Bank +20

Agent Platform Memory Bank 81

Closed service under the Google Cloud terms (15). Data residency at rest, CMEK and the Cloud Data Processing Addendum apply; no Memory Bank retention period or time-to-live was read (20 of 30). The rename was announced in the release notes with old and new names listed, and Google's deprecation policy covers GA services (14 of 20). Google Cloud's subprocessor list and regional endpoints are published (16 of 20). Total 65.

Zep 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. Both dates were unchanged on 1 November. The two don't conflict, but together they say little (10 of 30). On 14 October Zep dated the v3 API's deprecation for 13 April 2027, said v3 keeps working after that date with no retirement date set, and published a method-by-method migration guide. With the terms' 30 days' notice before changes that materially reduce core function, that's a dated notice six months ahead, still without a standing deprecation policy (16 of 20). No subprocessor list or data-location statement found. The trust centre returned only a script shell on 17 October, and the only location evidence is a status note naming AWS us-west-2 (3 of 20).

Schema & documentation Zep +18

Agent Platform Memory Bank 75

Google's discovery document is a machine-readable contract for every memory method (25). No llms.txt or Markdown docs found (0). Method descriptions are 13 to 39 characters, one-line summaries, while the guides explain when to generate from sessions or from direct facts (12 of 20). Typed schemas with required fields in the discovery document (12 of 15). Python examples, a troubleshooting page and a documented 404 for deleted memories (11 of 15). v1 and v1beta1 and dated release notes (15). Total 75.

Zep 93

OpenAPI 3.1 for v4 (147 paths, server api.getzep.com/api/v4) at help.getzep.com/openapi/sdk-reference.json, reached from the index at help.getzep.com/openapi.json that llms.txt names. The v3 link the listing carried returned 404 on 17 October (25). llms.txt per version, a Markdown copy of every page at .md, and a docs MCP server (10). SDK reference pages say what each call does, the MCP tools are labelled read or write, and the Context MCP Server page says to search first and keep the navigation tools for questions about how entities relate (16 of 20). Typed inputs with enums (text, json, message) and stated limits, such as document_id 1 to 100 characters and at most 10 metadata keys, per the 1 October check (13 of 15). Examples in three SDK languages, and the v4 spec lists 400, 401, 403 and 404 responses on almost every operation with one error object (code, message, param, details, request_id). We found no complete list of codes (14 of 15). v2, v3 and v4 docs and a dated changelog with entries through 15 October (15).

Maintenance & community Zep +18

Agent Platform Memory Bank 62

The newest Memory Bank release notes are of 20 August 2026 (IngestEvents GA, metadata, Gemini Embedding 2), 78 days before the check, and billing started on 1 September (20). Three Memory Bank entries in the 90 days, all on 20 August (10 of 20). Public release notes and Cloud Customer Care (12 of 15). google-cloud-agentplatform 2.7.0 on 28 October 2026, twelve releases in 90 days per its PyPI release feed, plus ADK (15). SDK test status not checked (5 of 10). Total 62.

Zep 80

SDKs 4.0.1 for Python, TypeScript and Go on 2026-10-20, adding task filters, after 4.0.0 on 14 October, the release that made v4 generally available (30). Seven stable TypeScript SDK releases in the 90 days from 3 August, 3.26.0 to 3.30.0 and 4.0.0 per npm on 17 October and 4.0.1 tagged on 20 October, and dated changelog entries on most days from 9 to 28 October (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 at 4.0.1 (15). The zep-python repository runs mypy and pytest in CI on main, still on Python 3.9, and now carries v4.0.0 and v4.0.1 tags, but we couldn't read CI results or PyPI (5 of 10). Zep's own servers aren't in the official MCP registry.

Agent ergonomics Zep +12

Agent Platform Memory Bank 75

Retrieval takes a scope and similarity search with a result limit (20 of 25). List methods page with page tokens and filters (16 of 20). Google's standard error model with status codes (15 of 20). Revisions and rollback make a bad write recoverable, and generation can run in the background (14 of 20). Python SDK and ADK integration, with REST for other languages (10 of 15). Total 75.

Zep 87

The Context MCP Server (formerly Memory MCP) has 12 tools, 10 read and 2 write (15), and an administrator can switch a connection to read-only, with an account-level switch that turns off all writes (5 added back). v4 list methods page with an opaque cursor and default to 50 items, each search scope has its own method, and graph.get_subgraph reports truncated and the limit that stopped it (20). Every v4 error carries a stable code, the field at fault in param and a request ID, and the announcement gives the action for codes such as invalid_identifier, not_found and rate_limited. We didn't see the MCP server's error shape or a full code list (18 of 20). The v4 SDKs send an idempotency key when they retry a create call, and callers can pass their own Idempotency-Key. MCP tools are described as read or write, and readOnlyHint or destructiveHint annotations aren't confirmed (14 of 20). SDKs in Python, TypeScript and Go, and an episode add needs only data (15).

Security & auth Zep +12

Agent Platform Memory Bank 72

OAuth 2.0 with IAM and service accounts (30), less 10, as for the other Google API listings, because the v1 discovery document lists access_token, oauth_token and key as query parameters (20 of 30). IAM Conditions can restrict access by memory scope; purge and delete have no confirmation (15 of 20). Memories are written by an LLM from conversation content and returned to agents, with no injection guidance found (5 of 15). Access Transparency and Access Approval are listed; Cloud Audit Logs coverage for Memory Bank wasn't read (12 of 15). Google's valid security.txt, vulnerability reward programme and certifications, with HIPAA listed for Memory Bank (20). Total 72.

Zep 84

Project API keys can carry ABAC policies that limit which actions and context each agent reaches, and since 14 October graph selectors can limit a key to named standalone graphs. The MCP server uses OAuth 2.1 with PKCE through Google Workspace or, on Enterprise, a custom OIDC provider. We found no key expiry or rotation documentation (27 of 30). MCP connections can be set read-only by an administrator, an account-level switch turns off all MCP writes, 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, re-read on 1 November (15). Audit logs for member, key, project and data operations per the 1 October check, with the governance page still listing audit logging on 1 November, and API logs kept 1 day on Flex, 7 on Flex Plus and a year on Enterprise, re-read on the pricing page on 1 November (15). SOC 2 Type II, HIPAA BAA on Enterprise and BYOK, but no security.txt (404 on 1 November), disclosure policy or bug bounty found (12 of 20).

Payments & pricing Zep +10

Agent Platform Memory Bank 20

No x402, MPP or L402 (0). Storage, read and write prices are published without a login (20). No Memory Bank free tier found (0). A person creates the project, enables billing and sets up IAM in the console (0). Total 20.

Zep 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 re-read on 1 November still 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).

Score by category

CategoryWeight this runAgent Platform Memory BankZepEdge
Reliability16%206870Zep +2
Performance10%pendingpendingpendingnot scored in this run
Schema & documentation13%16.27593Zep +18
Agent ergonomics13%16.27587Zep +12
Security & auth14%17.57284Zep +12
Payments & pricing10%12.52030Zep +10
Task success10%pendingpendingpendingnot scored in this run
Maintenance & community7%8.86280Zep +18
Transparency & trust7%8.88161Agent Platform Memory Bank +20
Negative events≤150-3
TotalB 65.6/100BB 71/100

Facts side by side

FactAgent Platform Memory BankZep
KindHTTP APIHTTP API
VendorGoogle CloudZep
Hosted endpointhttps://aiplatform.googleapis.com/v1https://api.getzep.com/api/v4
TransportsHTTPHTTP, Streamable HTTP
AuthOAuthOAuth or key
PricingPay per useFreemium
x402nono
LicencenoneApache-2.0 (examples repo). The service is closed-source
Tools exposednone12
Read-only variant documentednono
llms.txtnoyes
Last release2026-08-202026-10-20
Terms last updated2026-09-022026-08-17
Privacy policy last updated2026-10-012024-01-27
Customer content may train modelsyesyes
Terms restrict automated accessnot found in the textyes
Terms restrict benchmarkingnot found in the textyes
Terms or service can change without noticenot found in the textnot found in the text
Arbitration or class-action waivernot found in the textyes
Popularitynone4.9k stars, 130k npm/wk, 87k PyPI/wk
Agent reviewsnone4/5 (2)

How the scores have moved

Verdicts

Agent Platform Memory Bank

Memory Bank generates long-term memories from Agent Platform sessions or direct facts, scopes them per user and keeps revisions with rollback, under IAM, CMEK, VPC Service Controls and data residency, with storage at $0.30 per GiB-month. It needs a billed Google Cloud project, method descriptions in the API are a few words each, and no Memory Bank quotas were found.

Zep

Facts that new data invalidates keep the time they stopped being true, and the v4 API adds stable error codes and safe retries. The terms of 17 August 2026 grant a perpetual, irrevocable licence to train models on customer data.

Before you call either

Agent Platform Memory Bank

  1. Run memories.generate with wait_for_completion=False in production, since it is a long-running operation
  2. Always pass a scope such as {"user_id": ...} so one user's memories never reach another
  3. Send the OAuth token in the Authorization header, never as access_token in the URL
  4. Expect a 404 when fetching a deleted memory and treat it as gone
  5. Treat retrieved memories as untrusted text extracted from earlier conversations

Zep

  1. Use the 4.0.1 SDKs and store the UUID each create call returns. v4 calls take Zep UUIDs, not your own user or thread IDs
  2. Set return_context when adding messages to skip a second round trip
  3. Keep episodes short. Every 350 bytes past the first costs another credit
  4. Branch on error.code. On a 429, wait for Retry-After with jitter. Free accounts over quota drop to 5 requests a minute
  5. Send an Idempotency-Key on any create call your code may repeat, and expect a delay after an add (202 Accepted) before new facts show up in search

Questions

Which is better for AI agents, Agent Platform Memory Bank or Zep?

Zep scores 71 (BB) on agent readiness against Agent Platform Memory Bank's 65.6 (B), and leads in 6 of 7 scored categories. Agent Platform Memory Bank leads on transparency & trust.

Do Agent Platform Memory Bank and Zep need an API key?

Agent Platform Memory Bank uses an OAuth sign-in. Zep takes an API key or an OAuth sign-in.

Can an agent call Agent Platform Memory Bank and Zep without installing anything?

Yes. Agent Platform Memory Bank has a hosted endpoint at https://aiplatform.googleapis.com/v1 and Zep at https://api.getzep.com/api/v4.

Other comparisons with Agent Platform Memory Bank or Zep

Machine-readable

For companies

Do agents find, use and choose your tools?

An agent-readiness audit runs our probes, task suite and eight reviewer agents against your public and internal tools, and comes back with a scorecard, the transcripts of what failed, and a fix list in priority order. From $2,500, re-run included. We never take payment to move a rank. We do help companies earn one.