- Home
- Skills
- Agents & Orchestration
- codeagent memory
Works with the AI Agents you already use
codeagent memory
Local code graph and solution memory for symbol-level navigation and persistent project knowledge.
Free
codeagent memory
Example session with this skill installed
How did we solve the race condition in the auth flow last week, and which functions call the login handler now?
- Read your context and instructions
- Compiled the codeagent memory
I have queried the project memory and code graph to retrieve the previous solution and current call stack for the authentication flow.
- Memory: Hit #M-402 (reused).
- Code: Read
src/auth/handler.rs/login_handler. References: callers (in) forlogin_handler. - Write-back: nothing new to store.
Project Context and Code Graph
| Resource Type | ID/Path | Details |
|---|---|---|
| Memory Hit | M-402 | Summary: Resolved race condition in JWT validation by implementing a Mutex on the session store. |
| Symbol | src/auth/handler.rs/login_handler | Primary entry point for user authentication requests. |
| Callers (In) | src/router.rs/app_router | Routes POST requests to the login handler. |
| Callers (In) | src/middleware/session.rs/refresh_task | Triggers login logic during session expiry. |
Symbol Definition: login_handler
pub async fn login_handler(
State(state): State<Arc<AppState>>,
Json(payload): Json<LoginRequest>,
) -> Result<impl IntoResponse, AuthError> {
let mut session = state.store.lock().await;
// Implementation follows...
}
Next steps
- Verify if the
Muteximplementation inM-402requires upgrading to a fair semaphore for high-concurrency loads. - Run
cam ref src/auth/handler.rs/login_handler --dir outto see downstream dependencies. - Update the documentation in
docs/auth.mdto reflect the locked session store pattern.
Connects securely to your tools. The creator never sees your data.
About this skill
Context switching in large codebases leads to cognitive load and lost solutions. This skill manages a local SQLite code graph and project memory to track how problems were solved and how symbols relate across the project. It allows for precise symbol retrieval and relationship mapping without reading entire files into context.
What it does
- Persistent memory stores and recalls previous solutions or architectural decisions within the project.
- Symbol-level reading retrieves specific function or class definitions without loading full source files.
- Reference mapping tracks callers and callees to understand cross-file dependencies and impact.
- Graph indexing builds a local database of symbols, files, and relationships for fast querying.
- Local storage keeps all data within a
.cam/directory, ensuring no project code leaves the local environment.
How it works
- Recall checks for existing solutions or facts by turning the current task into a query.
- Index/Sync builds or updates the local graph when the filesystem changes.
- Traverse identifies symbols and follows their call chains to build context.
- Store commits new architectural facts or solved bugs to the project memory for future reuse.
Frameworks & tools
This skill requires the cam CLI tool installed on the system path and uses a local SQLite database stored in the project root. It integrates with agents via MCP or standard shell execution.
Why this beats prompting it yourself
Standard LLMs lack persistence across sessions and often struggle to navigate deep call stacks in large repositories. This skill provides a structured method for an agent to maintain its own "long-term memory" of your specific codebase and navigate it via a graph rather than simple grep searches.
Use cases
- Querying how a specific error was handled in a previous sprint.
- Finding all callers of a deprecated function before refactoring.
- Reading a specific symbol's implementation without hitting token limits on large files.
- Storing a complex integration logic summary for the next subagent to use.
Known limitations
Requires the cam CLI to be pre-installed on the host machine. Operates only on local project data and does not support remote database hosting.
Requires cam CLI pre-installed. Only supports local SQLite storage in .cam/ directory. Does not automate tool installation.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before any payment.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
1 use
Times used to date
Free forever
No account required to browse
Trust & safety
Security scan
Security scan passed 24 days ago
- Free with an Agensi account