release: autter 2.4.0 — repository memory before code, autter context and mcp connect

CHANGELOG.md

@@ -65,6 +65,20 @@ This changelog is automatically generated by AI when releases are published.
 
 For detailed design rationale and implementation notes, see `docs/superpowers/specs/2026-04-29-telemetry-streams-design.md`.
 
+## [2.4.0] - 2026-11-03
+
+### Added
+
+- **Repository memory before code**: `autter install` adds an `autter memory-hook` next to the checkpoint hooks of Claude Code, Codex, Gemini CLI, Droid, GitHub Copilot and Cursor, and plugins for OpenCode, Amp and Pi. The agent gets the repository's shared memory when the session starts, the sections that match each prompt, and the area page for each file it reads or edits. The hook gives up after 3 seconds and prints nothing on any failure. Turn it off with `autter config set feature_flags.memory_context false`. See [changelog/memory-before-code.md](changelog/memory-before-code.md).
+- **`autter memory`**: `autter memory search "<question>"` ranks the repository's memory, learnings, production lessons, team rules, past findings, wiki and code by meaning. `autter memory for <file>...` prints the area pages for the files you're about to edit, and `show`, `areas` and `ledger` read the rest. Add `--json` for scripts and agents. See [changelog/repository-memory-search.md](changelog/repository-memory-search.md).
+- **`autter context "<task>"`**: everything Autter knows that matters for a task in one call: memory facts, wiki pages and code structure, ranked and cut to a token budget, with the fact ids to cite. It returns the same answer as the `get_context` MCP tool, for agents that aren't connected to it. See [changelog/agent-context-and-mcp-connect.md](changelog/agent-context-and-mcp-connect.md).
+- **`autter memory skills install`**: writes the repository's own memory skill and a Cursor rule into the checkout, generated from its memory facts, so every teammate's agent has the memory with or without the CLI. `autter memory skills preview` prints them and writes nothing.
+- **`autter mcp connect`** and **`autter mcp disconnect`**: add or remove Autter's MCP server in the user config of Claude Code, Codex, Cursor, VS Code, Gemini CLI and OpenCode. Only the URL is written, and the server signs you in with OAuth the first time an agent uses it. Every file is backed up before it changes.
+
+### Changed
+
+- **Repository memory skill**: the skill moved to `autter-repo-context` in [Autter-dev/autter-skills](https://github.com/Autter-dev/autter-skills), so agents without the CLI's hooks get it too. It tells agents to call `autter context` first, or `get_context` when the MCP server is connected, and to cite the fact ids they rely on.
+
 ## [2.3.0] - 2026-10-25
 
 ### Added

Cargo.toml

@@ -5,7 +5,7 @@ resolver = "3"
 
 [package]
 name = "autter"
-version = "2.3.0"
+version = "2.4.0"
 edition = "2024"
 default-run = "autter"
 

npm/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@autter/cli",
-  "version": "2.3.0",
+  "version": "2.4.0",
   "description": "Git-native AI authorship tracking: records which lines were written by AI agents and links them to the agent, model, and prompt. This package is a thin bootstrapper that installs the native autter binary.",
   "license": "Apache-2.0",
   "homepage": "https://autter.dev",