Codeledger
Give Claude Code or Cursor stable, session-to-session project memory so the agent picks the right files and context instead of re-guessing every chat.
Overview
Codeledger is a Build-phase MCP server that provides deterministic context selection and local compounding memory for AI coding agents.
What is this MCP server?
- Deterministic context selection so agents load the same relevant slices of the repo across sessions
- Local memory store that compounds instead of relying only on the model’s short context window
- stdio MCP via npx @codeledger/cli mcp start (npm @codeledger/cli v0.10.7)
- Targets AI coding agents that need repeatable file and decision recall on solo codebases
- No cloud dependency implied in the catalog description—memory stays on your machine
- Server version 0.10.7 in server.json
- npm package @codeledger/cli with stdio transport
- Launched via npx with positional args mcp and start
Community signal: 8 GitHub stars.
What problem does it solve?
Agent chats forget prior decisions and reload the wrong files, wasting tokens and reintroducing mistakes across sessions.
Who is it for?
Indie developers on long-lived repos who run Claude Code or Cursor daily and want local, deterministic memory without building a custom RAG stack.
Skip if: Teams that need shared org-wide knowledge graphs, audited cloud memory, or one-off prompts with no MCP setup.
What do I get? / Deliverables
After you register the server, your agent can pull consistent, session-spanning context from local memory instead of starting cold each time.
- Registered stdio MCP server in your agent config
- Local compounding memory usable across coding sessions
- More consistent context selection for repo-aware agent turns
Recommended MCP Servers
Journey fit
Canonical shelf is Build → agent-tooling because the server exists to wire persistent memory into coding agents during implementation. Agent-tooling is where MCP memory and context-selection servers belong in Skillselion’s journey map.
How it compares
MCP memory and context-selection integration, not a markdown planning or brainstorming skill.
Common Questions / FAQ
Who is Codeledger for?
Solo and indie builders using MCP-capable coding agents who want local memory and predictable context loading on their own machine.
When should I use Codeledger?
Use it when multi-session agent work on the same repo keeps losing track of files, conventions, or prior fixes and you want that state to compound locally.
How do I add Codeledger to my agent?
Add a stdio MCP entry that runs npx with @codeledger/cli and the mcp start positional arguments, then restart Claude Code, Cursor, or your compatible client.