Somewhat related, I built ccrider [1], which indexes the session transcripts that agents already write to disk (Claude Code, Codex, Copilot) into a local SQLite FTS database. It has a TUI, CLI, and an MCP server so the agent can search past sessions itself. Same local-first idea, based the raw logs rather than a maintained digest. Should work nicely with recall as well.
Cool project. I like the neuroscience analogy with decay and consolidation.
I've been working on a related problem from the other direction: Claude Code and Codex already persist full session transcripts, but there's no good way to search across them. So I built ccrider (https://github.com/neilberkman/ccrider). It indexes existing sessions into SQLite FTS5 and exposes an MCP server so agents can query their own conversation history without a separate memory layer. Basically treating it as a retrieval problem rather than a storage problem.
I get the concerns, but there's a flip side nobody's talking about: what about experienced developers who had years of ideas bottled up and never had the bandwidth?
I used to think open source was something I'd focus on when I retired. Instead I'm doing it now. I built Clippy (https://github.com/neilberkman/clippy) from scratch, a much better bcopy, that fixes a real gap in pbcopy, and it hit 200 stars and Homebrew core in seven months. Neither of these would exist without AI assisted development. My DocuSign Elixir SDK sat basically unchanged from 2021 to 2025, other than a few external contributions. I knew it needed a full migration of the OpenAI code gen layer, LiveBook examples and a million other tweaks. With coding agents I implemented everything useful I could imagine.
Am I the only one? There must be other devs out there resurrecting stale projects or shipping tools they never had time for.
That makes resuming easier, if you remember to rename in your current session. But that doesn’t solve being able to search history from earlier in your current session or from your entire history of sessions. Or resuming sessions for which you’ve forgotten the name.
Nice! Tangentially related: I built a (MacOS only) tool called clippy to be a much better pbcopy. It was just added to homebrew core. Among other things, it auto-detects when you want files as references so they paste into GUI apps as uploads, not bytes.
clippy image.png # then paste into Slack, etc. as upload
clippy -r # copy most recent download
pasty # copy file in Finder, then paste actual file here
Interesting - this is the exact problem José Valim cited when creating Elixir: he was working on Rails multi-core performance in 2010 and found that "writing multi-core software, which is software that runs on all cores with Ruby, was not really straightforward."[1] Ruby's GIL meant only one thread executes at a time.
Fleury's arriving at similar conclusions from the C/systems side: make multi-core the default, not an opt-in. Though his approach still requires explicit coordination (LaneIdx(), barriers, range distribution) vs BEAM where the scheduler automatically distributes processes across cores.
Different tradeoffs for different domains, but both are reacting to the same backwards reality where we program single-core-first on multi-core hardware.
Inspired by the copy-paste point in this post, I added agent buffer tools to clippy, a macOS utility I maintain which includes an MCP server that interacts with the system clipboard. In this case it was more appropriate to use a private buffer instead. With the tools I just added, the server reads file bytes directly - your agent never generates the copied content as tokens. Three operations:
buffer_copy: Copy specific line ranges from files to agent's private buffer
buffer_paste: Insert/append/replace those exact bytes in target files
buffer_list: See what's currently buffered
So the agent can say "copying lines 50-75 from auth.py" and the MCP server handles the actual file I/O. No token generation, no hallucination, byte-for-byte accurate. Doesn't touch your system clipboard either.
The MCP server already included tools to copy AI-generated content to your system clipboard - useful for "write a Python script and copy it" workflows.
(Clippy's main / original purpose is improving on macOS pbcopy - it copies file references instead of just file contents, so you can paste actual files into Slack/email/etc from the terminal.)
Don't know why this is getting downvoted. Agree with this so hard, as a continually aggrieved Fly customer (close to becoming an ex-customer). The too cool for school schtick gets old fast when they don't have the goods to back it up.
I tried Terragon on an open source repo I maintain and ended up with a PR: "Showing 11,643 changed files with 2,441,792 additions and 584 deletions." Yikes!
Thanks. I just released a new version with UTI based MIME type detection as the primary method with a fallback to the existing content-based detection.
clippy is a bit different from these - rather than requiring manual flags for MIME types, it automatically detects the content type. The idea is to make the CLI's copy/paste feel as intuitive as the GUI's.
I might tackle Linux (and Windows) at some point, but contributions are definitely welcome!
Shannon: Fast local search for your Claude conversation history
Claude's web/desktop search is pretty lacking - no boolean operators, no date filtering, no advanced search syntax. So I built a fast local search tool for exported conversations (not API - this is for your regular Claude.ai chats). Supports AND/OR/NOT operators, wildcard search, date ranges, sender filtering, and more.
I think the TUI mode is particularly useful - you can navigate between conversations, jump back to Claude.ai, and copy/save artifacts directly.
I tried to make the whole process of generating exports and importing them incrementally as painless as possible. It auto-discovers export files and handles incremental updates.