Acontext, Inc.
Acontext is an open-source context-data and skill-memory platform for AI agents. It captures agent sessions and outcomes, stores context and artifacts, observes task execution, and turns successful work into reusable, human-readable Markdown skills so agents can improve across runs.
At a Glance
- AI application developers
- Enterprise software and enterprise AI teams
- Teams building coding agents and autonomous agents
- Developers needing context engineering, agent observability, and self-learning infrastructure
- +1 more
AI Tools by Acontext, Inc.
(1)
Acontext
Agent Skill Memory Layer
Discussions
No discussions yet
Be the first to start a discussion about Acontext, Inc.
Latest News
ACONTEXT completed a seed round of millions of dollars, jointly invested by Sequoia China Seed Fund and Mingshi Capital.
Agent Skills as a Memory Layer: Acontext described its shift from storing what was said to capturing what worked and writing reusable Markdown skills.
Context Engineering Tips for Agents That Run for Hours, Not Seconds.
From Demo to Production: Why Agent Context Starts to Break.
Products & Services
Managed cloud service with dashboard, project and session management, collaboration, monitoring, and observability without requiring users to maintain infrastructure.
Self-hostable context-data platform for storing multimodal messages and artifacts, observing agent work, tracking tasks, and learning reusable skills.
Automatically distills validated task outcomes into reusable, human-readable Markdown skill files organized in learning spaces; skills can be read, edited, exported, and reused across agents and frameworks.
APIs and SDKs for sessions, messages, multimodal context, artifact disks, context editing at retrieval time, task tracking, and sandbox execution.
Market Position
Acontext positions itself as a context-data and skill-memory layer rather than an agent framework, database, or conventional observability product. It differentiates from conversation-memory products such as Mem0 and Zep by learning from what an agent did and how the task turned out, and from RAG/vector stores by storing portable Markdown skills and using tool-based, progressive disclosure. It complements frameworks such as LangGraph and Agno and observability tools such as LangSmith and Langfuse rather than replacing them.
Leadership
Founders
Ye Jianbai (Gus Ye)
Founder and CEO. He participated in development of the Bing search advertising algorithm at Microsoft Research Asia, later worked on distributed large-model training at Beyond Light Years, and worked on model training and RAG systems at Muyan Zhiyu. He led the open-source Memobase project before Acontext.
Executive Team
Ye Jianbai (Gus Ye)
Founder and CEO
Former Microsoft Research Asia researcher on Bing search advertising; later worked on distributed large-model training at Beyond Light Years and model training/RAG at Muyan Zhiyu; led Memobase before Acontext.
Founding Story
The team started from the observation that agents work in one run but fail to remember what worked, why they failed, or how to reuse successful execution patterns. Ye Jianbai chose context as the entrepreneurial starting point after seeing that agent-generated data is far larger and more operationally important than chatbot data, with the initial vision of infrastructure for storing, monitoring, and learning from agent context. Acontext was then launched to turn context and execution experience into an agent-development data platform.
Business Model
Revenue Model
Acontext offers a managed cloud platform alongside a free/open-source, self-hostable platform. The public materials describe cloud access, API/SDK usage, and infrastructure-managed services, but do not publish a detailed commercial billing model.
Pricing Tiers
Apache License 2.0 repository; users run the stack on their own infrastructure with the Acontext CLI and Docker.
Target Markets
- AI application developers
- Enterprise software and enterprise AI teams
- Teams building coding agents and autonomous agents
- Developers needing context engineering, agent observability, and self-learning infrastructure
- Open-source and self-hosting adopters
- Building production AI agents and long-running agent workflows
- Coding agents that need to reuse fixes, deployment procedures, and operational experience
- Multimodal chatbots and applications handling messages, images, audio, documents, and artifacts
- Context engineering for sessions that grow over time
- Agent observability, task auditing, and success/failure analysis
- Personalized, user-specific skill libraries and self-improving agents