Grep-style search for your AI chat history

Your best answers are buried
in chat history. Grep them back.

ChatGrep indexes your ChatGPT, Claude, Gemini, Cursor, Codex, and Aider conversations into one local, searchable archive. Regex or semantic search, jump links to the source, Markdown export — and an MCP server so your agents can recall past chats mid-task.

Local-first index. Nothing leaves your machine.

Everything you need to search your AI history

Purpose-built retrieval for AI power users, developers, and agent fleets whose best work is trapped in months of chat logs.

One index for every AI tool

Point ChatGrep at your ChatGPT, Claude, Gemini, Cursor, Codex CLI, and Aider exports — plus local agent logs. One command builds a single searchable archive.

Regex + semantic search

Match exactly like code when you know the words, or search by meaning when you only remember the idea. Millisecond results across thousands of chats.

Jump links to the source

Every result shows the app, the timestamp, and a link back to the exact message — so you can reopen the original conversation in one click.

Export to Markdown/JSON

Turn any search into a curated doc. Ship answers to your repo, wiki, or prompt library instead of re-asking the model.

MCP server for agents

Expose your archive to Claude, Cursor, or any MCP client. Your agents grep past conversations as a tool call and reuse what you already figured out.

Local-first privacy

The index lives on your machine. Nothing is uploaded, nothing is trained on. Team sync is opt-in, per workspace, encrypted end to end.

From buried to searchable in one command

Index once, search everywhere — in your terminal, your editor, and your agents.

01

Point it at your exports

Run `chatgrep index` against ChatGPT, Claude, and Gemini data exports, Cursor/Codex/Aider local logs — or any folder of chat JSON. One command builds your local index.

02

Search like code

Regex when you know the words, semantic when you only remember the idea. Results show the app, the timestamp, and a jump link straight to the source message.

03

Recall via MCP

Add the chatgrep MCP server to Claude, Cursor, or your agent stack. Your agents grep past conversations mid-task — yesterday’s perfect answer becomes one tool call.

Simple pricing

Start free for 14 days. Pay only when your archive grows.

Starter

$19/mo

For power users with months of personal chat history.

  • 1 machine, 100k messages indexed
  • Regex + semantic search
  • Jump links to source conversations
  • Markdown/JSON export
Most popular

Team

$49/mo

For teams who never want to re-ask what someone already asked.

  • 5 seats, 1M messages
  • Shared team workspace
  • MCP server included
  • Priority re-indexing
  • Slack + webhook sync

Business

$99/mo

For agent fleets where recall is part of the workflow.

  • Unlimited machines + seats
  • SSO + audit log
  • Fleet-wide MCP access
  • 1-year retention
  • Priority support

Loved by AI power users

Teams and solo builders use ChatGrep to turn months of AI conversations into an on-demand knowledge base.

I found a prompt I wrote eight months ago in about four seconds. ChatGrep paid for itself before lunch.

Sarah Chen
CTO, PromptLayer

Our support agents recall past AI answers through the MCP server instead of re-asking. Consistency across the team went way up.

Marcus Webb
Head of Engineering, SynthAPI

Regex when I know the words, semantic when I don’t. It’s the grep I always wanted for chat logs.

Elena Ruiz
Co-founder, VectorScale

Questions & answers

Everything else you might want to know before you start grepping.

ChatGrep builds one local, searchable index across your AI chat histories — ChatGPT, Claude, Gemini, Cursor, Codex CLI, and Aider exports, plus local agent logs. Search every message with regex or semantics, jump back to the source, and export results to Markdown or JSON.
ChatGPT, Claude, Gemini, Cursor, Codex CLI, and Aider today. Anything you can point a folder at — chat JSON exports, CLI logs, local app storage — works too.
No. The index is local-first: nothing is uploaded and nothing is used for training. Team workspace sync is opt-in and encrypted.
In-app search covers one tool at a time — and often only conversation titles. ChatGrep searches message text across every tool you use, works offline, and returns jump links plus app and timestamp for every hit.
A local MCP server that exposes your archive as a search tool. Claude, Cursor, or your own agents can grep past conversations mid-task — so yesterday’s answer is one tool call away.
Yes — one command turns any search into a Markdown or JSON doc, ready for your repo, wiki, or prompt library.

Ready to grep your AI history?

Join power users turning months of AI conversations into a searchable, local knowledge base — for themselves and their agents.