skaft-software/octet

This page introduces an external open-source repository. It is not an HDATF product.

What it does

octet is a terminal coding agent for developers who want a native Rust runtime with cloud or local models. It takes prompts, source files and optional media and returns answers, file edits and command results. One process manages the agent loop, sessions, policies, persistence and terminal interface. Tools can be added through subprocess extensions, while MCP and subagents are optional integrations. Its first Pi-compatible release supports parts of the Pi extension ecosystem, but compatibility has documented gaps and arbitrary Pi transcripts are not imported. Sessions use append-only JSONL records with branches and checkpoints, and interrupted mutating calls are not automatically replayed. Model credentials and the chosen provider or local model must be prepared, and Windows support is still a preview. Full operating-system access is the default, and its approval-based safe mode is not a sandbox for commands or trusted extensions.

License

MIT Permissive. Commercial use and changes are allowed if the copyright notice is kept.

More in this category

• metadist/synaplan-desktop
  Synaplan Desktop connects a user's computer to a Synaplan workspace so that chat can lead to local file work. It takes requests, permitted files and installed Agent Skills and produces local artifacts such as reports, tables, calendar files and unsent email drafts. The native client uses Tauri and Vue and communicates with the server through pairing APIs and an Anthropic-compatible Messages gateway. A scoped pairing key is normally kept in the operating system's secret store. Local execution uses approved interpreters and structured program arguments, rejecting shells, network command-line tools and inline code. Skills needing unrestricted shell or network access do not run unchanged, and local Python or other required tools must already be available. The release is an unsigned preview for Windows, macOS and Linux, with signed public installers not yet published according to the README. Source inspection also shows that Windows timeout handling terminates the direct child rather than guaranteeing termination of its descendants, so isolation claims require separate validation.
• darkmatter/nixmac
  nixmac is a macOS configuration manager that places an AI agent in front of Nix and nix-darwin settings. It is intended for Mac users who want to describe package or system-setting changes instead of editing Nix files directly. A natural-language request and an existing configuration become edited files, a change summary, and, through a separate apply path, a new system configuration. The agent reads files, searches packages and documentation, makes edits, and revises them using check results. The inspected implementation uses a dry-run Nix check during evolution, which does not perform the full package build described more loosely by the README. The apply path performs an actual build as the user and then activates the resulting system with administrative privileges. Git history and saved system paths support review and rollback, but restoring files and reactivating a system are separate operations. The documented environment is macOS with Nix flakes, nix-darwin configuration, and devenv, with models supplied through OpenRouter, OpenAI-compatible endpoints, or Ollama. Dry-run success does not establish successful activation or compliance with the user's intent, and a separate review button does not by itself prove an enforced backend approval boundary.
• agentclientprotocol/codex-acp
  This adapter connects Codex to editors and clients supporting ACP. It is a protocol server, not a separate language model or a complete development environment. It accepts text, images and workspace context, starts the Codex App Server, and returns replies, tool events, file changes and permission requests. Clients can select models, reasoning effort, sandbox modes and allowed user permission profiles. Native subagent sessions and some mid-turn capabilities require capability negotiation with the client. The npm package includes a compatible Codex dependency, but client launch configuration and Codex authentication remain separate prerequisites. Authentication supports ChatGPT login, API keys and negotiated custom gateways. Installation does not automatically configure every editor or establish a safe approval policy, and some permission-profile fields use an experimental schema.
• govijr/terminal-mac
  This tool gives computer-use AI agents a separate virtual Mac. Apple Virtualization runs macOS, while a helper inside the guest captures the screen and handles input. Agents request screenshots, clicks, drags, keystrokes and status through MCP or the command line. MCP is a common communication method connecting AI tools to external functions. Requests travel as messages through a shared folder, and results return as images or structured responses. Users can view the same guest in a compatible terminal or a native virtual machine window. Getting started requires an Apple silicon Mac with macOS 14 or later, a macOS restore image and storage space. A person must create the guest account and grant Screen Recording and Accessibility permissions during initial setup. The guest does not inherit host apps or signed-in sessions, but default networking can reach the host or local network. It is an early release candidate, and official validation documentation still lists end-to-end guest installation and screen control as incomplete.
• deeplethe/sandstorm
  Sandstorm manages isolated execution environments for agent services that spend time waiting between requests. Developers supply an application template, resource limits, and sandbox creation requests. It returns sandbox state and command output through an HTTP API and provides a management console. A master keeps desired state in SQLite and distributes work to nodes. Idle environments move from running to paused, disk snapshots, and finally a state retaining only their volume. The node gateway restores an environment when traffic arrives, subject to admission and restore queue limits. It supports gVisor containers and Firecracker microVMs, with separate network namespaces for sandboxes. Setup requires Linux with cgroup v2 and backend dependencies, and Firecracker also requires a compatible virtualization environment. Snapshot placement depends on backend and host compatibility, and capacity pressure can produce refused requests rather than guaranteed immediate execution.

Only repositories in the ranked Trendshift lists are included, and the lists are used only to find candidates. We do not copy their ranks. Descriptions, licenses and star counts come from each GitHub repository. The notes are our own reading. We have not tested these projects, and a place on a trending list does not prove quality.