assets

scale, there are generally three big problems that emerge: 1. Agents are long-running compared to other parts of your application. Extremely long-running processes are tricky because deploying new infra or hitting request timeouts on serverless runtimes will interrupt their execution. 2. They are stateful: they generally store internal state which governs the next step in the execution path 3. They require access to lots of fresh data, which can either be queried during agent execution or needs to be continuously refreshed from a data source. (These problems are more specific to agents which execute remotely -- locally running agents generally don't have these problems) Pickaxe is designed to solve these issues by providing a simple API which wraps durable execution infrastructure for agents. Durable execution is a way of automatically checkpointing the state of a process, so that if the process fails, it can automatically be replayed from the checkpoint, rather than starting over from the beginning. This model is also particularly useful when your agent needs to wait for an external event or human review in order to continue execution. To support this pattern, Pickaxe uses a Hatchet feature called `waitFor` which durably registers a listener for an event, which means that even if the agent isn't actively listening for the event, it is guaranteed to be processed by Hatchet and stored in the execution history and resume processing. This infrastructure is powered by what is essentially a linear event log, which stores the entire execution history of an agent in a Postgres database managed by Hatchet. Full docs are here: https://pickaxe.hatchet.run/ We'd greatly appreciate any feedback you have and hope you get the chance to try out Pickaxe.

rked so well that we decided to streamline it and make it more compact and easy to install. In the process, we found that using Rust and the MCP Rust SDK makes it really compelling to build MCP servers, with significant performance improvements: - 40x faster startup time (~50ms vs ~2s) - 70% less memory usage (~15MB vs ~50MB) - Single 3MB binary vs ~200MB Node.js dependencies - No subprocess overhead or external dependencies As Andy put it [4]: "Terminals are one of the oldest and most prolific UI frameworks in all of computing. And they are stateful." Without HT, an agent struggles to manage this state directly; with HT-MCP, an agent can observe and interact with the terminal just like a human does. This is an early first release and we'd love to hear if you find it useful. Here is a demo using it from Claude Code[2] and Memex[3] (our own desktop agent coding tool). Install it via Homebrew: `brew tap memextech/tap && brew install ht-mcp` 1. https://github.com/andyk/ht by Andy Konwinski 2. https://github.com/user-attachments/assets/e70a3240-77f5-4ef... 3. https://github.com/user-attachments/assets/6a1b6e76-5d5c-4ba... 4. https://news.ycombinator.com/item?id=40552257

e it tools to fetch a directory tree, get the content of any file or folder, diff any two branches/commits/tags, and pull context from issues/PRs. This works surprisingly well, even on a massive repo like VS Code's (>27M tokens). Ironically, I only recently learned that GitHub has its own official server for this! Luckily, my version has some capabilities that theirs doesn't (yet): 1. Get a clean directory tree: It uses the Git tree object to create a readable tree structure, which helps the AI (and humans) understand the project layout faster. 2. Get a diff between any two points: This isn't limited to PRs; you can compare any two branches, commits, or tags. 3. Get raw repo contents: It can fetch file contents as plain text or folder contents in the tree format, ready for AI to process directly. Any feedback is highly appreciated! Try it out: https://github.com/baonguyen09/github-second-brain Demo Video: https://github.com/user-attachments/assets/ecb05256-8bb0-427...