Knock

Knock is a local-first code map and memory tool that gives AI coding agents persistent context over codebases.

Knock provides a local code map and memory system for AI coding agents like Claude Code, served over the Model Context Protocol (MCP). It indexes a repository without running code in it, writing short notes on modules, setup documentation, dependencies, and past decisions. When an agent asks a question, knock returns a small, ranked answer with file and line pointers instead of forcing the agent to read entire files. Agents can also save findings using a remember tool, and knock automatically flags notes as stale if the underlying code changes. It supports command-line operations for tasks like searching knowledge and checking index freshness.

Knock at a glance

Best for: Software engineer
Runs on: CLI, API
Pricing: Free
Released: November 2026
Category: AI coding tool
Website: github.com/hasangilani/knock

Key features

Builds local code maps with exact line ranges
Stores agent findings and decisions with the remember tool
Flags outdated notes when code changes
Provides command-line tools for searching knowledge and checking status
Serves context over the Model Context Protocol
Automatically indexes changed files

Use it for

Finding code locations and architectural context for AI coding agents
Saving and retrieving bug hunt findings across agent sessions
Querying deployment steps and project decisions
Checking repository index freshness and status

Alternatives to Knock

Cost X-Ray — It prices usage below the tool level by analyzing raw API request bytes.
Claude Context — It reduces token costs and provides deep context for large codebases.
Claude Tap — Inspect prompts, tool calls, and token usage locally without a hosted dashboard.
Edgee Mod for Claude Code — Manage models and track token cost savings inside your terminal conversation.
Oh My ClaudeCode — Orchestrates parallel Claude, Gemini, and Codex CLI workers.
Claude Code Local — Runs Claude Code completely on-device without cloud limits or API keys.
Token Beggar — It reduces cost and quota usage on large tasks while keeping quality similar.
Claude Code Workflows — Provides pre-built, automated workflows for Claude Code development processes.
Claude Boot Beats — Adds music and repository stats to every coding session start.
Compactly — Compacts sessions at natural breaks and shows turn costs and cache expiration.
Reviewer Agent — It uses a separate AI model to review code and prevent premature agent stopping.
claude-mem — It shares persistent, searchable agent memory across sessions and devices.

All Knock alternatives