Skip to main content
PowerContext slogan: Not only memory but a full story. When work changes hands, chat history is a poor place to keep decisions. PowerContext stores the evidence, decisions, and handoff state so the next person or agent can continue. It runs as its own context service. It is not a chat transcript store. You begin a task, an agent moves it forward, and you or another agent continues later. Context stays with the work. PowerContext 101 covers working paths and integration boundaries. For install options, complete interfaces, and deployment, use the official PowerContext docs. Start with one result you can check: write a Memory, find it, then get a cited PreparedContext. That loop proves the Server and Scope before a framework, host hook, or model provider adds another failure boundary.
Finish this first loop first. For complete install options, interface details, and deployment, use the latest published official docs.

​
A 30-second mental model

PowerContext runs three paths at once. They may cite one another. They do not collapse into one lifecycle. Keep knowledge. A only proves that something happened. Write a when you will reuse it. The next request recalls it through a . Review first. and managed first become a . A model may draft it. Successful generation is not approval. decides what stays. An approved Skill still needs an explicit export. A handoff can stay temporary. A is packed for the next actor. It becomes a durable milestone only after an explicit commit. Understand PowerContext in 30 seconds: keep knowledge, review first, and hand off work as three separate paths. lasts for one request. It includes Citations and is labeled as untrusted historical evidence. An agent may use it as evidence. The text cannot grant tools or replace current system and user instructions.

​
Continue by goal

Separate the objects first

Source, Memory, PreparedContext, and Handoff are easy to blur together. Their boundaries make the API and host integrations much easier to reason about.

Run the common APIs

Start a local Server, then use the Python Client to complete the first loop: write a Memory, find it with search_memory, and use prepare_context to obtain a cited PreparedContext.

Connect a Python agent

Compare where LangGraph, LangChain, and Pydantic AI connect, along with their maturity and capability differences.

Configure a common agent

Install, activate, and verify Hermes, OpenClaw, WorkBuddy, or Agent Plugin. Credentials, durable writes, and Review decisions still require your authorization.

Configure a coding agent

See how Codex, Claude Code, DSH, OpenCode, and Pi map the same Server contract onto different host lifecycles.

Follow the learning path

You do not need to read the whole site first. Pick one path that matches your work, finish it, then come back. If you are already stuck, start with Troubleshooting.
If PowerContext fails while automatically fetching context, keep working. If you explicitly write a Memory, commit a Handoff, approve a proposal, or export a Skill, you must see the failure. Read fail-open for the full rule.
Not sure which card to open? Start with the learning path. If you are already stuck, match the symptom on Troubleshooting.
oceanbase/powercontext
Loading repository data...