screenpipe vs Zep (2026)

screenpipe

by Mediar, Inc.

Zep

by Zep AI

Pros

• ✓Local-first by architecture rather than by policy: capture, storage and search all sit on your own machine, cloud model calls send only the query, and Ollama support means the whole loop can run without a single cloud call.
• ✓A first-party MCP server rather than a claim: npx -y screenpipe-mcp exposes search-content and export-video over a local REST API on port 3030, with a one-click install into Claude Desktop from the app's own settings panel and documented stdio setups for Claude Code, OpenAI Codex, Cursor and Warp.
• ✓Nothing about the price or the setup is gated: all three tiers publish a rate with no quote-only step and the currency sits in the site's own structured data, and getting running is a desktop install on macOS, Windows or Linux against a documented five-minute quickstart.

Limitations

• ⚠The trust portal shows SOC 2 Type 2, ISO 27001, GDPR and HIPAA all as COMPLIANT with no in-progress qualifier anywhere on the page, and the site's own security page does not corroborate it, so we publish only SOC 2 Type II, GDPR and CCPA. There is little independent evidence to check the vendor against either: 21,150 GitHub stars but only two Product Hunt reviews and no G2 profile.
• ⚠Section 4 of the terms has you consent, by using the service, to the vendor creating Deidentified Data and Analytics Data from your usage and using it to improve and train its models. Customer content, output and personal information are excluded without separate written agreement, but the deidentified pathway is consent-by-use rather than opt-in.
• ⚠Public source is not an open-source licence: commercial use requires a paid licence under the Screenpipe Commercial License, and GitHub reports the repository licence as Other. Running it also costs roughly 5 to 10 GB of local disk per month, and on macOS the app captures nothing until screen recording and accessibility access are granted in System Settings.

Pros

• ✓Temporal knowledge graph automatically tracks how facts change over time. When information is updated or contradicted, Zep invalidates outdated facts while preserving historical context, making it more accurate than vector-only memory systems for complex multi-session agent workflows.
• ✓Benchmark-leading retrieval with 80.32% accuracy at 189ms on LoCoMo and 63.8% on LongMemEval versus Mem0's 49%. Delivers measurably better temporal reasoning at lower latency using single-shot retrieval with no slow agentic tool-call loops.
• ✓Graphiti MCP server gives Claude Desktop, Cursor, and other MCP clients persistent knowledge graph memory across sessions. Python, TypeScript, and Go SDKs with webhook support and no agent framework lock-in, backed by Fortune 500 customers including AWS and Samsung.

Limitations

• ⚠Steep pricing jump from free tier (10,000 credits per month for prototyping) to Flex at $125 per month ($104 annual) with no intermediate tier. Early-stage teams that outgrow free credit limits face a significant cost increase with no middle-ground option.
• ⚠Credit-based pricing makes costs difficult to forecast at production scale. Episodes consume 1 credit per 350 bytes with auto-topup charges at 20% of monthly allocation, requiring careful per-request credit monitoring to avoid unexpected billing.
• ⚠Self-hosted Graphiti requires Docker, FalkorDB or Neo4j, and LLM API key management. Teams seeking a fully managed zero-infrastructure MCP memory solution face meaningful setup and operational overhead before reaching production.