To solve this, I built AntiGoldfishMode: a CLI tool that gives your AI assistant a persistent, local-only memory of your codebase. There are enough cloud-based tools that solves some of the issues relating to AI persistent memory, but not a lot that combines all the features I have placed in AGM.
- Verifiable Zero-Egress - How to verify: Run agm prove-offline\n- Supply Chain Integrity for Shared Context: The .agmctx Bundle, Checksums First, Cryptographic Signature: An Ed25519 key pair (generated and stored locally in keys) is used to sign the SHA-256 hash of the concatenated checksums. This signature is stored in signature.bin.\n- Policy-Driven Operation\n- Transparent Auditing via Receipts and Journal. You should never have to wonder what the tool or your AI coding agent did. It is like a "Glass Box" where you see and verify every move your AI coding agent makes, every edit is recorded.
Receipts: Every significant command (export, import, index-code, etc.) generates a JSON receipt in receipts. This receipt contains a cryptographic hash of the inputs and outputs, timing data, and a summary of the operation.\nJournal: A journal.jsonl file provides a chronological, append-only log of every command executed and its corresponding receipt ID. This gives you a complete, verifiable audit trail of all actions performed by the tool.\nThis combination of features is designed to provide a tool that is not only powerful but also transparent, verifiable, and secure enough for the most sensitive development environments.
It's built with a few core principles in mind:
Local-First & Air-Gapped: All data is stored on your machine. The tool is designed to work entirely offline, and you can prove it with the agm prove-offline command.\nTraceable & Verifiable: Every action is logged, and all context exports can be cryptographically signed and checksummed, so you can verify the integrity of your data.\nNo Telemetry: The tool doesn't collect any usage data.\nThe core features are MIT-licensed and free to use. There are also some honor-system "Pro" features for advanced code analysis and stricter security controls, which are aimed at professional developers and teams.
You can check out the source code on GitHub: https://github.com/jahboukie/antigoldfish
I'd love to hear your feedback"},"title":{"fullyHighlighted":false,"matchLevel":"partial","matchedWords":["ai","assistant"],"value":"Show HN: An open-source,security first,local-first memory tool for AI assistants"}},"_tags":["story","author_Jahboukie","story_44888862","show_hn"],"author":"Jahboukie","created_at":"2025-08-13T14:24:24Z","created_at_i":1755095064,"num_comments":0,"objectID":"44888862","points":1,"story_id":44888862,"story_text":"I've been using AI coding assistants like Copilot and Claude a lot, but I constantly hit the limits of their context windows, forcing me to re-explain my code over and over. I also work on projects with sensitive IP, so sending code to a third-party service is a non-starter.
To solve this, I built AntiGoldfishMode: a CLI tool that gives your AI assistant a persistent, local-only memory of your codebase. There are enough cloud-based tools that solves some of the issues relating to AI persistent memory, but not a lot that combines all the features I have placed in AGM.
- Verifiable Zero-Egress - How to verify: Run agm prove-offline\n- Supply Chain Integrity for Shared Context: The .agmctx Bundle, Checksums First, Cryptographic Signature: An Ed25519 key pair (generated and stored locally in keys) is used to sign the SHA-256 hash of the concatenated checksums. This signature is stored in signature.bin.\n- Policy-Driven Operation\n- Transparent Auditing via Receipts and Journal. You should never have to wonder what the tool or your AI coding agent did. It is like a "Glass Box" where you see and verify every move your AI coding agent makes, every edit is recorded.
Receipts: Every significant command (export, import, index-code, etc.) generates a JSON receipt in receipts. This receipt contains a cryptographic hash of the inputs and outputs, timing data, and a summary of the operation.\nJournal: A journal.jsonl file provides a chronological, append-only log of every command executed and its corresponding receipt ID. This gives you a complete, verifiable audit trail of all actions performed by the tool.\nThis combination of features is designed to provide a tool that is not only powerful but also transparent, verifiable, and secure enough for the most sensitive development environments.
It's built with a few core principles in mind:
Local-First & Air-Gapped: All data is stored on your machine. The tool is designed to work entirely offline, and you can prove it with the agm prove-offline command.\nTraceable & Verifiable: Every action is logged, and all context exports can be cryptographically signed and checksummed, so you can verify the integrity of your data.\nNo Telemetry: The tool doesn't collect any usage data.\nThe core features are MIT-licensed and free to use. There are also some honor-system "Pro" features for advanced code analysis and stricter security controls, which are aimed at professional developers and teams.
You can check out the source code on GitHub: https://github.com/jahboukie/antigoldfish
I'd love to hear your feedback","title":"Show HN: An open-source,security first,local-first memory tool for AI assistants","updated_at":"2026-03-05T22:34:49Z"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"hamzakat"},"comment_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["air","gapped","codebase","ai","assistant"],"value":"Hi there
I worked on this tool in an afternoon (with GLM 4.6) to solve a problem I'm facing: in some workspaces, AI coding assistants aren't allowed, forcing you to use web-based chatbots or local LLMs . To get meaningful help, you need to provide context about your codebase, but manually copying files is tedious.
While there are great tools like Gitingest and Repomix, they require installing packages which isn't straightforward in air-gapped work environments! I wanted something more portable - just a shell script that works anywhere you have a terminal.
GitHub: https://github.com/hamzakat/ingest.sh"},"story_title":{"fullyHighlighted":false,"matchLevel":"partial","matchedWords":["codebase"],"value":"Ingest.sh \u2013 A portable shell script for codebase summarization"},"story_url":{"matchLevel":"none","matchedWords":[],"value":"https://github.com/hamzakat/ingest.sh"}},"_tags":["comment","author_hamzakat","story_45580207"],"author":"hamzakat","comment_text":"Hi there
I worked on this tool in an afternoon (with GLM 4.6) to solve a problem I'm facing: in some workspaces, AI coding assistants aren't allowed, forcing you to use web-based chatbots or local LLMs . To get meaningful help, you need to provide context about your codebase, but manually copying files is tedious.
While there are great tools like Gitingest and Repomix, they require installing packages which isn't straightforward in air-gapped work environments! I wanted something more portable - just a shell script that works anywhere you have a terminal.
GitHub: https://github.com/hamzakat/ingest.sh","created_at":"2025-10-14T14:08:22Z","created_at_i":1760450902,"objectID":"45580208","parent_id":45580207,"story_id":45580207,"story_title":"Ingest.sh \u2013 A portable shell script for codebase summarization","story_url":"https://github.com/hamzakat/ingest.sh","updated_at":"2026-03-05T22:52:23Z"}],"hitsPerPage":20,"nbHits":2,"nbPages":1,"page":0,"params":"query=air-gapped+codebase+AI+assistant&advancedSyntax=true&analyticsTags=backend","processingTimeMS":91,"processingTimingsMS":{"_request":{"queue":1,"roundTrip":17},"fetch":{"query":77,"scanning":12,"total":90},"total":91},"query":"air-gapped codebase AI assistant","serverTimeMS":92}