OpenAI Agents SDK Memory
https://openai.github.io/openai-agents-python/ref/memory/
OpenAI's agent SDK. Sessions for working context within an agent loop; RunContextWrapper for structured persistent state. April 2026 update added categorised long-term memory: project / user / policy. Long-term tier typically backed by external vector DBs (Pinecone, etc.).
Active · Verified 2026-10-22 · fresh as of 2026-10-22 · 52 cited facts, 1 with a third-party source · 1 recorded change · Suggest a correction
At a glance
- Tier
- T2
- Section
- Framework-embedded memory
- Latest release
- v0.24.0 (2026-10-15) registry fix
- GitHub
- openai/openai-agents-python — ~28.6k stars repo fix
- Pricing
- SDK OSS free; pay for OpenAI API model usage at standard token pricing (e.g. GPT-5.4: $2.50/$15 per M tokens) vendor fix
- Funding
- OpenAI raised $40B+ total funding across multiple rounds; $300B+ valuation; no independent funding for SDK vendor fix
- Freshness
- 2026-10-22 fix
- Layer
- agent-infra repo fix
- Licensing
- open-source registry fix
- SDK languages
- python, typescript repo fix
- Protocols
- mcp-client vendor fix
- Developer activity
- OpenRank 22.9 (12-mo avg 29.19) · 153 participants — 2026-09 third-party fix
Memory taxonomy
- storage
- kv
- retrieval
- injection
- persistence
- cross-session
- update
- overwrite
- unit
- episode
- governance
- inspectable
- conflict
- overwrite
When to use
Optimised for: OpenAI-native agent SDK with Sessions + categorised memory
Anti-fit: not for non-OpenAI providers (provider-coupled)
Pros & cons
Pros
First-party from OpenAI — fastest path if you're building exclusively on OpenAI models.
Cons
Vendor-locked; less feature surface than dedicated memory layers.
Claims & capabilities
Sessions-based working-context layer over Responses API; SDK handles context length, history, and continuity automatically. Storage backends: SQLite (default), SQLAlchemy (production), Dapr, OpenAI Conversations API. Compaction via OpenAIResponsesCompactionSession. April 2026 added categorised long-term memory (project / user / policy)
Technical surface
- API surface
- REST, SDK: Python, JS/TS
- Backend storage
- custom (OpenAI-managed)
- Deployment
- Both (OSS library self-hosted; relies on OpenAI or Azure OpenAI API endpoints; no dedicated SDK hosting)
- Embedding model
- multiple supported
- Multi-tenancy
- hard-isolation
- MCP
- native (first-party) — Agents SDK MCP support
- A2A
- searched not found
- OpenTelemetry
- via adapter — OpenAI traces + OTel exporters
Alternatives to OpenAI Agents SDK Memory → live options ranked by shared design, licence, hosting and MCP support
Similar systems
Other framework-embedded memory in the catalog, ranked by inbound references.
- Spring AI ChatMemory T3
Wraps conversation memory behind an Advisor abstraction. MessageWindowChatMemory default 20-msg sliding window (preserves system messages); pluggable ChatMemoryRepository backends (in-memory, JDBC, Cassandra, Neo4j). Memory injected as cross-cutting concern at the ChatClient level.
- Flowise Memory T1
Discontinued: Flowise (acquired by Workday, Aug 2025) was sunset in 2026 — code freeze 2026-07-29, GitHub repo archived 2026-08-13, end of life 2026-08-31. Was a visual LangChain canvas with LangChain-native memory nodes (Buffer, Buffer Window, Conversation Summary Buffer) backed by an internal chat_message DB table; Buffer Window kept the last K interactions, Summary Buffer compressed older turns while keeping recent ones verbatim.
- LangGraph Persistence T2
Distinct from LangMem. Built-in checkpointer saves graph state per super-step (short-term, thread-scoped). Store System adds long-term hierarchical key-value memory across threads with optional vector search + TTL. Postgres / Mongo / Redis stores for production.
- AutoGen Memory T2
AutoGen's memory layer: ListMemory chronological context plus teachable agents that vectorise corrections; integrates with Mem0/Zep rather than building deep memory natively. Microsoft AutoGen is in maintenance mode (community managed, bug and security fixes only), and Microsoft points new projects to Microsoft Agent Framework.
- CrewAI Memory T2
Memory subsystem inside the CrewAI orchestration framework; integrates with Mem0 for the long-term tier.
- Hermes Agent memory (Nous Research) T3
Nous Research's open-source self-improving agent keeps an agent-curated MEMORY.md/USER.md notebook injected at session start, full-text (FTS5) search over past sessions that returns raw messages without LLM summarisation, and optional Honcho user modelling. one of the most-starred agent projects of 2026.
Sources (51 cited facts, 1 third-party)
Each fact links to the page it came from; the badge says who published it (how to read sources).
- What it is Sessions + RunContextWrapper + categorised long-term vendor
- What it does & what's distinct OpenAI's agent SDK. Sessions for working context within an agent loop; RunContextWrapper… vendor
- Claims / benefits Sessions-based working-context layer over Responses API; SDK handles context length, his… vendor
- Modalities text vendor
- Created 2025-03 (launched March 2025 as production-ready evolution of Swarm) vendor
- Latest release v0.24.0 (2026-10-15) registry
- License MIT registry
- GitHub openai/openai-agents-python — ~28.6k stars repo
- Funding OpenAI raised $40B+ total funding across multiple rounds; $300B+ valuation; no independe… vendor
- Pricing SDK OSS free; pay for OpenAI API model usage at standard token pricing (e.g. GPT-5.4: $2… vendor
- Compliance OpenAI enterprise plans: SOC 2 Type II, GDPR, HIPAA BAA available; Azure OpenAI adds gov… vendor
- Data handling Enterprise/API: data not used for training by default; encrypted in transit/at rest vendor
- Deployment Both (OSS library self-hosted; relies on OpenAI or Azure OpenAI API endpoints; no dedica… vendor
- HQ US (San Francisco, CA) vendor
- Founders / pedigree OpenAI; corporate product (Sam Altman CEO; Greg Brockman; ex-OpenAI researchers) vendor
- Memory volume / scale no scale claim vendor
- API surface REST, SDK: Python, JS/TS vendor
- Backend storage custom (OpenAI-managed) vendor
- Multi-tenancy hard-isolation vendor
- Encryption at-rest + in-transit vendor
- SSO / RBAC SSO + RBAC vendor
- Embedding model multiple supported vendor
- Namespace primitives per-tenant vendor
- Contradiction handling always overwrite (last-write-wins) vendor
- MCP support native (first-party) — Agents SDK MCP support vendor
- OTel via adapter — OpenAI traces + OTel exporters vendor
- Import / export JSON via Responses / threads API vendor
- Integration count 0 vendor
- Orchestration both (single Agent + handoffs/multi-agent runs; session-scoped memory) vendor
- Programmatic control full CRUD via Python/JS SDK vendor
- Pricing specifics SDK OSS free; pay OpenAI API (e.g. GPT-5.4 $2.50/$15 per M tokens) — per-token-stored vendor
- Agent abstraction ReAct (Sessions + RunContext) vendor
- Memory primitives multiple (Sessions working + categorised long-term project/user/policy + external vector) vendor
- LLM lock single-vendor (OpenAI; OpenAI-Compatible adapters possible) vendor
- Runtimes Python, Node/TypeScript repo
- Session handling sticky session + persistent thread vendor
- Validated verticals OpenAI enterprise + developers (horizontal) vendor
- Time to running <10min (pip install + OpenAI key) vendor
- Anti-fit not for non-OpenAI providers (provider-coupled) vendor
- Optimised for OpenAI-native agent SDK with Sessions + categorised memory vendor
- Adjacent infrastructure requires OpenAI API; BYO persistence layer vendor
- Pros First-party from OpenAI — fastest path if you're building exclusively on OpenAI models. vendor
- Cons Vendor-locked; less feature surface than dedicated memory layers. vendor
- Project & sources SDK home memory reference April 2026 update cookbook: long-term notes vendor
- architecture-deployment-locus local vendor
- licensing-bucket open-source registry
- positioning-platform-vs-product platform vendor
- taxonomy-three-layer-rollup agent-infra repo
- developer-activity-opendigger {"activity":434.56,"as_of":"2026-09","openrank_12mo":29.19,"openrank_current":22.9,"part… third-party
- sdk-languages python,typescript repo
- protocol-support mcp-client vendor
What changed
Changes to this record's lifecycle, licence, release, pricing and hosting facts, from the catalog's git history. Subscribe (Atom)
Related systems
References (2)
- OpenAI GPT family (GPT-5 / GPT-4o / o3 / o4) depends on at runtime — adjacent-infrastructure cell: requires OpenAI API; BYO persistence layer
- Pinecone builds on — Long-term tier typically backed by external vector DBs (Pinecone, etc.).