Acontext
Agent Skills as a Memory Layer
Acontext is an open-source skill memory layer for AI agents that automatically captures learnings from agent runs and stores them as readable, editable Markdown skill files. Instead of opaque embeddings, it represents agent memory as plain files that can be inspected, edited, and shared across different agents, LLMs, and frameworks like LangGraph, Claude, and AI SDK. The system extracts tasks from conversation sessions, distills successful patterns and failures, and writes them to structured skill files defined by your schema. Agents retrieve skills through tool calls (get_skill, get_skill_file) using progressive disclosure rather than semantic search, and all skills can be exported as ZIP archives for reuse anywhere without vendor lock-in.
Key Features
Use Cases
- 01Building agents that learn from mistakes and successes across multiple runs without losing context
- 02Creating reusable skill libraries that can be shared across different AI frameworks and LLMs
- 03Debugging agent behavior by inspecting readable Markdown memory files instead of opaque embeddings
- 04Personalizing agents that remember user preferences and working patterns stored as editable skill files
- 05Developing team agents that share learned skills through version-controlled skill repositories
- 06Migrating agent knowledge between platforms by exporting skills as portable ZIP archives
Related Skills
View moresuperpowers
An agentic skills framework & software development methodology that works.
skills
Skills for Real Engineers. Straight from my .agents directory.
skills
Public repository for Agent Skills
ponytail
Makes your AI agent think like the laziest senior dev in the room. The best code is the code you never wrote.
Acontext — FAQ
What is Acontext?+
Acontext is an open-source skill memory layer for AI agents that automatically captures learnings from agent runs and stores them as readable Markdown skill files. These files can be inspected, edited, and shared across different agents, LLMs, and frameworks without embeddings or vendor lock-in.
How do I install Acontext for my agent?+
Install the Python SDK with 'pip install acontext' or the TypeScript SDK with 'npm i @acontext/acontext'. Then initialize a client with your API key from acontext.io, create a learning space, attach sessions, and store messages to begin automatic skill learning.
Which AI frameworks and clients work with Acontext?+
Acontext works with any framework that reads files, including LangGraph, Claude (Code and Desktop), OpenAI Agent SDK, Vercel AI SDK, Agno, and smolagents. Skills are plain Markdown files that can be used with any LLM or agent platform.
Do I need an API key to use Acontext?+
For cloud-hosted use, you need an Acontext API key (starts with 'sk-ac') from acontext.io which includes free credits. For self-hosted deployment using Docker, you need an OpenAI API key (or compatible LLM) since the skill distillation agent requires tool-calling capabilities.
Is Acontext free to use?+
Acontext is open-source under Apache License 2.0 and free to self-host with Docker. The cloud-hosted version at acontext.io offers free credits to get started, with the option to scale as needed.
How does Acontext store and retrieve agent skills?+
Acontext automatically extracts tasks from agent conversation sessions, distills outcomes (successes and failures) using an LLM, and writes structured Markdown files following your SKILL.md schema. Agents retrieve skills through tool calls like get_skill rather than semantic search, enabling progressive disclosure and reasoning-based retrieval.
How do I install Acontext?+
Open the source repository on GitHub and follow its README. Acontext is a skill — MCP Agents Market links you directly to the official repo.
Is Acontext free?+
Acontext is an open-source project hosted on GitHub. Check the repository for its license and any usage requirements.