Which package? This is EverOS Cloud — the managed SaaS client (pip install everos-cloud).

Want to self-host? Use the open-source everos library instead.

EverOS Cloud — Python SDK

Give your AI agents memory that persists across sessions — managed, searchable, and typed. Add a conversation; EverOS turns it into structured, retrievable memory you can query in one call.

Why EverOS Cloud

• Self-evolving memory — memory doesn't just pile up, it improves. Background consolidation merges related episodes and refines user profiles over time, so recall gets sharper the more your agent is used.
• Structured memory, not chat logs — extracts episodes, user profiles, and reusable agent cases & skills from raw conversations, so retrieval returns meaning, not transcripts.
• Retrieval that fits the query — keyword, vector, hybrid (default), or agentic multi-step search, plus Context Retrieval for questions that need several rounds of searching.
• Knowledge bases — ingest documents into a searchable topic library alongside conversational memory; ingest is async and reports progress through the task API.
• Multimodal — attach images, audio, and documents to any message.
• Built for production — fully managed (no vector DB or extraction pipeline to run), with low-latency retrieval and high-concurrency throughput. The engineering guarantees you don't get from self-hosting.
• Fully typed (pydantic v2) — every request/response is a typed model with full hints, so you get editor autocomplete and validation instead of raw dicts.

Install

pip install everos-cloud

Pre-releases need --pre: pip install --pre everos-cloud.

Upgrading from the 0.4.x client? 1.x is a rewrite with a new API surface — see the migration guide. Pin everos-cloud<1 to stay on the old client.

Quickstart

Get an API key from the EverOS Console, then:

from everos_cloud import EverOS

with EverOS(api_key="sk-...") as client:
    client.add(session_id="session-1", messages=[
        {"sender_id": "user-1", "role": "user", "content": "I love hiking in the mountains"},
    ])

    results = client.search("outdoor hobbies", user_id="user-1")
    print(results)

Knowledge bases work the same way — ingest is asynchronous, so wait on the task:

kb   = client.kb_create("Employee Handbook")
ack  = client.doc_ingest(kb.id, "Leave policy", "Employees accrue 20 days...")
task = client.task_wait(ack.task_id)          # polls until the document is queryable

hits = client.kb_search(kb.id, "how much leave do I get")

Full usage — every memory operation, knowledge base, async task, profile editing, and multimodal upload — is in quickstart.md.

Two ways to call the API

EverOS covers the common calls with plain kwargs in and the response's .data out. Memory methods are bare verbs named after their route (add, search, get, flush, edit, delete, update, delete_by_ids, feedback, context); every other resource is <resource>_<verb> (kb_create, doc_ingest, task_wait, tag_bind), so typing client.kb lists the knowledge-base surface.

Everything the API offers — all 37 operations, including knowledge-base categories and document topics — is on the generated typed clients, reachable as client.memory, client.storage, client.knowledge, client.tasks. Those take and return the full typed models, so responses arrive as an envelope you read .data from. The per-endpoint reference for them is under docs/.

client.kb_create("Handbook")                          # facade   -> KbData
client.knowledge.create_knowledge_base({"name": "…"})  # generated -> envelope, .data
client.knowledge.list_topics(kb_id, doc_id)            # generated only

Documentation

• Full SDK usage
• Quickstart
• API Reference
• Core Concepts

back to top

EverMind Ecosystem

EverMind connects memory research, production-ready products, and practical integrations into one open-source ecosystem.

Together, these projects form EverMind's research-to-runtime stack: methods and benchmarks become reusable memory infrastructure, products, and agent integrations.