Save token
Create a new entry for the token, API key, or other secret you want KeyLore to manage. The raw value is stored separately from the metadata shown to AI clients.
Core workflow
The UI is where you define the boundary between secret material and AI-visible context.
A useful KeyLore setup gives agents enough metadata to choose correctly while keeping raw tokens out of their view.
Write context
Write metadata that explains what service the credential belongs to, what it should be used for, and any scope or environment distinctions that matter.
Preview what the AI sees
Check the metadata as if you were the agent. It should be descriptive enough to support correct selection, but it should not contain the secret itself or unnecessary sensitive detail.
Test a client workflow
Use a supported AI coding tool or MCP client to confirm that it can find the credential through metadata and request brokered access when it needs to perform the task.
Writing guidance
What makes metadata useful to an agent.
- Prefer metadata that states purpose and scope plainly, such as which API or environment a credential is for.
- Avoid vague labels that force the agent to infer intent from abbreviations or internal naming habits.
- Treat metadata as part of the interface contract with the AI tool: concise, accurate, and free of raw secret values.
Next step
Connect MCP
Use the client-specific connection flow after the token passes its brokered test.
Common question
Migrating from .env
See the shortest safe path for moving existing tokens into KeyLore.