perplexityai/SPACE

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

What it does

SPACE is Perplexity's Apache-2.0 platform, written in Rust, that runs agent code isolated in Firecracker microVMs (lightweight virtual machines). It is for applications that need a separate execution environment for an agent, a user workspace or a code-running job, and the application decides what to run and what to keep. A template is the starting environment, and a sandbox is an instance of it that runs several commands in its own VM and writable filesystem. Applications create sandboxes and command sessions through an HTTP API and poll an operation ID until each sandbox lifecycle change completes, and command output and the exit result arrive as server-sent events over a separate connection with a session-scoped token. Pausing keeps memory on the worker and suspending saves a snapshot, and a sandbox suspends by default after six hours without activity. An E2B compatibility wrapper lets existing E2B clients use SPACE as their backend, and its tested baseline is the Python packages e2b 2.21.1 and e2b-code-interpreter 2.6.2. Network policies allow or deny the destinations a sandbox can reach, and an optional credential rule, off by default, has the TLS-intercepting gateway add a credential header to requests for a named host. The local setup builds a Kind (Kubernetes in Docker) cluster on an apt-based Linux x86_64 host with KVM, and macOS is outside the supported setup. The repository description speaks of pluggable runtimes, while the documented implementation currently covers Firecracker on KVM only. Version 0.8.0 is the public baseline, breaking changes remain possible until 1.0, and only the latest minor release receives security fixes.

License

Apache-2.0 Permissive, with a patent grant. Commercial use is allowed; keep the notices and state your changes.

More in this category

• SakanaAI/AI-Scientist-v3
  The Agentic Scientist, in the repository AI-Scientist-v3, is Sakana AI's Apache-2.0 Python pipeline in which coding agents carry a task through experiments to a paper. It is for computational research with a program to improve and an evaluator that scores it, and the examples are circle packing on a CPU and GRPO training on two GPUs. Input is a task directory with task.md, an initial program and an evaluator, and drafting a task from a short brief alone is experimental, works for small tasks such as circle packing and may fail on larger ones. Six phases run in order: task drafting, investigation that repeats ideas and experiments, ablations against baselines, a sealed final evaluation, post-hoc analysis and paper writing. Each run creates a directory with the task snapshot, attempts and results, cost records and a LaTeX paper with its PDF, and a run inspector shows it in the browser. The paper phase links each claim to evidence files with hashes and gathers references from Semantic Scholar, OpenAlex or Google Scholar, and it fails when citations or evidence links do not pass validation. It drives Claude Code, Codex, Cursor, OpenCode, Gemini and Pi through the Headless CLI, and an optional review phase sends the finished paper to an OpenAI-based reviewer model. Getting started needs Python 3.11 or later with uv, Docker, and API keys or a Claude login in .env, GPUs are needed only for GPU presets, and agsci doctor is the first check. The README warns that agents run any command with full tool permissions, inherit the API keys and are steered by the files they read, and it asks for runs inside Docker. One GRPO evaluation trains for one to two hours, and the cost files separate provider-reported costs from list-price estimates, leave hardware costs out and are run telemetry, with provider invoices as the billing record.
• wangmiaozero/pi-harness
  Pi-Harness is an AGPL-3.0-only Electron desktop app that wraps Pi Coding Agent in a visual workspace while the native Pi runtime still runs the agent. It runs on macOS, Windows and Linux, and sessions, configuration, tools, Skills and extensions stay compatible with Pi, so work can continue in Pi outside the app. The workspace covers conversations, projects, a file browser and editor, Git diffs, history, commits and worktrees, model providers, and Skills, packages and MCP. Its Harness view records trace, replay, a run tree of forks and retries, tokens, cost and duration, and adds tool policies, per-run budgets and checkpoints. In the source, evaluation is computed from run evidence such as test, lint, typecheck and build commands with their exit codes and the Git state, with no AI judge. The default policy denies file changes outside the workspace and force pushes, asks before deletes, pushes and network commands, confirms commands such as rm -rf, and leaves budget limits empty. Agent teams assign tasks with dependencies and priorities, run them in order or in parallel, pass results on, use review gates and keep changes apart in Git worktrees. A history view reads the local conversation records of Claude Code, Codex, Cursor, Cursor Agent, Gemini CLI, OpenCode and Pi read-only and continues a task in a new Pi chat through a handoff. Installers are on GitHub Releases, the latest being 2.0.0 of 14 October, which follows Pi 1.2.0's breaking changes, and macOS community builds may be unsigned. On Windows and Linux, API keys are encrypted in the app's own vault, while the value written into Pi's models.json is the plain key.
• onehumanbeing/esp32-iphone-agent
  ESP32 iPhone Agent is an MIT-licensed project made of a Python server and Arduino firmware that lets an AI vision model operate a real iPhone without a Mac, Xcode, a developer account, an app on the phone or a jailbreak. It is for developers and hobbyists who want to automate an iPhone through the same paths a person uses, the screen and a pointer, and the README asks users to automate only devices and accounts they own or are allowed to use and to follow each app's terms. The phone mirrors its screen over AirPlay to UxPlay on a macOS or Linux computer, where GStreamer turns it into JPEG frames, and an ESP32-S3 board plugged into the phone's USB port appears as a standard USB mouse and keyboard. With iOS AssistiveTouch (an accessibility setting that shows an on-screen pointer) turned on, a click becomes a tap and a drag a swipe, and the board carries out only simple one-line commands such as click and type received over 2.4 GHz Wi-Fi. All decisions are made by a FastAPI server on the computer, and in that code the web API and the board port listen on every network interface by default without a login, with commands sent as plain text over TCP. A user types a goal into the web page on port 8890, a planner model writes a plan of up to 6 steps, and an actor model returns up to 6 actions such as tap and swipe in coordinates from 0 to 1000 for each new screenshot. Because the pointer moves relatively, the server calibrates it by wiggling it and finding it on the mirrored screen, checks after each action whether the screen changed and tells the model, and re-plans when the same action repeats without effect. A run ends at 40 steps or after 3 errors in a row, and the web page shows a step wall with the screenshot behind each decision and its action drawn on top, plus an emergency stop that the Esc key also triggers. Setup needs an ESP32-S3 board flashed with the firmware (the tested board is the Seeed XIAO ESP32S3 at about 7 dollars), a USB-C iPhone or a Lightning iPhone with a camera adapter, and Python 3.10 or later. The model can be any OpenAI-compatible chat API that accepts images, with separate models possible for the planner and the actor, and since every call carries the current screenshot, with a cloud API whatever the phone screen shows goes to that provider. Development and timing used a local model on two NVIDIA DGX Sparks (about 1.3 seconds per action and 2 seconds per plan), and the bundled serve folder holds an experimental server whose full agent run is still untested. The README lists its limits: iOS keeps passcode entry, Apple Pay and DRM video out of mirroring so the agent cannot see them, mirroring stops when the phone locks, typing is ASCII only, and landscape apps remain unsupported.
• T-Auto/dsh-ops
  dsh-ops is a third-party plugin for DeepSeek Harness (DSH) that gives AI agents on Windows their own bash and PowerShell 7 plus Rust file tools so they spend fewer tokens and less time on repeated PowerShell errors. Version 0.3.0 supports only Windows x64, targets DSH 0.2.0-rc.2 and needs Node.js 22.19 or later on the 22 line or version 24 and later, and it installs from the DSH plugin marketplace or with an npx command. The shell component, on by default, publishes ops_bash and makes the host pwsh tool use the bundled PowerShell 7 while leaving the system default terminal and PATH unchanged. The file component, also on by default, runs the file tools in a separate Rust worker and offers a dry run and atomic replacement for edits. Of the two components that are off by default, the background component handles long jobs only in danger-full-access sessions, and the PTC component adds ops_batch only in PTC-only sessions, in which tools are called through an SDK from code. The ops_batch tool runs copy, move and remove operations with SHA-256 checks, a per-item state written before each effect and resume after interruption, and its audit records are plain local records under the .dsh-ops folder that the workspace owner can alter. In the DSH read-only and workspace-write modes, arbitrary Bash and PTC code runs inside OS confinement with no unconfined fallback, but the Windows host reports this confinement as partial. The danger-full-access mode skips this confinement by design, and the plugin backs up the original permissions of workspace root folders owned by the current user and then adds WRITE_OWNER on that folder only. The bundled bash and PowerShell 7 arrive as optional npm packages of 161.2 MB and 107.6 MB packed, and if a registry mirror lacks them the install still ends with exit code 0 and no warning. The 97 percent and 65 percent token reductions in the README come from the author's own benchmark, while the PTC design document states that no fixed saving and no real model-turn benchmark is claimed. The LICENSE file puts the plugin under MIT and FastCtx under Apache-2.0, the bundled shells keep their GPL (Git for Windows) and MIT (PowerShell) terms, and PROVENANCE.md says the GPL source delivery has not been reviewed. The docs record that live installation on each DSH version has not been attested and that on 2026-10-08 an earlier version made every new DSH session fail through a tool registration conflict, which the current design avoids by registering only its own tools.
• mksglu/context-mode
  Context Mode is an MCP server and hook plugin for developers who use AI coding agents, and it keeps large tool output out of the model's context window and restores the working state after a conversation is compacted. Of its 11 MCP tools, ctx_execute, ctx_batch_execute and ctx_execute_file run code in 12 languages in a separate subprocess and pass only the printed result into context. The ctx_index, ctx_search and ctx_fetch_and_index tools store Markdown and fetched web pages in a SQLite FTS5 index and return only the matching passages ranked by BM25. Hooks redirect curl, wget and web fetch calls to these tools and record file edits, git operations, errors and user decisions in a per-project SQLite database, and the README says that starting a new session without --continue deletes the previous session data. After compaction they inject a resume snapshot with per-category summaries and search calls, and the code at the pinned commit ignores the 2 KB limit that the README states. It needs Node.js 22.5 or later or Bun, Claude Code installs it as a plugin from its marketplace, and other clients take MCP server and hook entries in their config files or use a client-specific plugin. An MCP-only install gives the tools without automatic routing, and session continuity is full on Claude Code and others, partial on Cursor, Codex CLI and Kiro, and absent on the Antigravity IDE and Zed. The README states that nothing leaves the machine and that there is no telemetry, but the server checks the npm registry for the latest version at startup and ctx_insight opens the Insight dashboard on context-mode.com. The hooks/platform-bridge.mjs file, which the README does not describe, posts every session event to the configured Insight server whenever platform.json holds an API key and a server address, after masking some secret patterns, email addresses and home folder paths. The ctx_execute_file tool is confined to the project folder, but the README says ctx_execute and ctx_batch_execute run arbitrary code with the process's own file access and asks users to keep host-level sandboxing on. The user's permission deny rules also apply inside the sandbox tools, and ctx_fetch_and_index allows local and private network addresses unless CTX_FETCH_STRICT=1 is set. The LICENSE file is the Elastic License 2.0 (ELv2), which allows use, modification and redistribution but forbids offering the software to third parties as a hosted or managed service and removing license notices.

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.