Acontext provides an open-source memory layer for AI agents, storing learned knowledge as agent skills in Markdown files with SDKs for Python and TypeScript. It includes a local/self-host option and a REST API backbone with a UI dashboard.
Collecting history — the radar snapshots this repo daily. The trend line appears after 3 days of data (1 so far).
What it is
Acontext is an open-source skill memory layer for AI agents. It automatically captures learnings from agent runs and stores them as agent skill files that can be read, edited, and shared across agents, LLMs, and frameworks.
How it works
- Store: Session messages lead to task outcomes, distillation via an LLM pass, and writing updates to skills following a SKILL.md schema.
- Recall: On next runs, skills are exposed to the agent context through get_skill/get_skill_file without embedding search.
- Memory uses agent skills as the memory representation and supports exporting skill files as ZIP.
Getting started
-
Install the Python SDK:
pip install acontext. -
Initialize client (cloud or self-hosted):
- Cloud:
client = AcontextClient(api_key=os.getenv("ACONTEXT_API_KEY"),) - Self-hosted:
client = AcontextClient(base_url="http://localhost:8029/api/v1", api_key="sk-ac-your-root-api-bearer-token",)
- Cloud:
-
Basic workflow:
- Create a learning space and session:
space = client.learning_spaces.create()session = client.sessions.create()client.learning_spaces.learn(space.id, session_id=session.id) - Store messages during agent run:
client.sessions.store_message(session.id, blob={"role": "user", "content": "My name is Gus"})client.sessions.store_message(session.id, blob={"role": "assistant", "content": "Hi Gus! How can I help you today?"}) - Wait for learning and list skill files:
client.learning_spaces.wait_for_learning(space.id, session_id=session.id)skills = client.learning_spaces.list_skills(space.id) - Download skills:
for skill in skills:client.skills.download(skill_id=skill.id, path=f"./skills/{skill.name}")
- Create a learning space and session:
-
Quickstart for self-host:
- Install CLI: see instructions in README; example command shown:
curl -fsSL https://install.acontext.io | shfollowed byacontext server upto start backend locally.
- Install CLI: see instructions in README; example command shown:
Recent releases
- ui/v0.1.14 (2026-04-08): UI v0.1.14 with dependency updates.
- sdk-ts/v0.1.21 (2026-04-08): TypeScript SDK v0.1.21 with dependency updates.
- package-claude-code/v0.1.3 (2026-04-08): Claude Code Plugin v0.1.3.
- core/v0.1.22 (2026-04-08): Core v0.1.22.
- cli/v0.1.21 (2026-04-08): CLI v0.1.21.
Full changelog links are provided in the release notes for each component.
Traction
- Stars: 3656
Behind the repo
- This repository is associated with memodb-io and focuses on memory layer for AI agents, with multiple submodules for UI, SDKs, core, and plugins.
Caveats
- License: Apache-2.0
- Created: 2025-07-16
- Last push: 2026-07-14
- Open issues: 36
- Language: JavaScript
- Licensing, open issues, age are listed as facts.