Space by Perplexity

Perplexity released SPACE (Sandboxed Platform for Agentic Code Execution), an open-source platform that runs code in isolated Firecracker microVMs and exposes an HTTP API for creating Linux environments, executing commands, collecting output, and managing sandbox lifecycles. SPACE uses templates as starting environments and sandboxes as per-instance VMs with writable filesystems, with asynchronous lifecycle operations that return an operation ID to poll and command output delivered over a separate data-plane connection using a session-scoped attachment token. The release includes an E2B Compatibility Wrapper for the tested E2B Python SDK, a Kubernetes-based local Kind deployment path, warm pools to skip cold boots, and hosted Guide and API Reference documentation.

SPACE Sandboxed Platform for Agentic Code Execution is a sandbox platform for running code in isolated Firecracker microVMs. Applications use its HTTP API to create Linux environments, execute commands, collect output, and manage sandbox lifecycles. Use SPACE when your application needs a separate execution environment for an agent, a user workspace, or a code-running job. Your application decides what to run and what results to keep; SPACE manages the sandbox, its processes, and the connection carrying command output. A template supplies the starting environment. A sandbox is an instance of that environment with its own VM and writable filesystem; multiple commands can run in the same sandbox. php Your application | +-- Create sandbox from template -- Wait for readiness | +-- Create command session -------- Attach to start the process | Read output and exit result | +-- Keep, pause, or suspend -------- Wake when needed | +-- Delete sandbox Sandbox lifecycle changes are asynchronous: the API returns an operation ID that you poll for completion. Command output uses a separate data-plane connection and a session-scoped attachment token. Follow Installation and configuration https://github.com/perplexityai/SPACE/blob/main/docs/installation.md to deploy SPACE in a local Kind cluster and authenticate. Then follow Your first sandbox https://github.com/perplexityai/SPACE/blob/main/docs/first-sandbox.md to run a command and collect its output. The setup guide covers Kubernetes; the architecture guide https://github.com/perplexityai/SPACE/blob/main/docs/architecture.md request-and-execution-flow also explains how SPACE can run without it. Read the hosted Guide https://space.docs.perplexity.ai/guide/index.html and API Reference https://space.docs.perplexity.ai/api/index.html online. Both update automatically after changes merge. The links below point to source documentation in this checkout. | Guide | What you will learn | |---|---| | Versioning and compatibility https://github.com/perplexityai/SPACE/blob/main/docs/versioning.md | Release progression, supported guests, service stability, and upgrade/rollback requirements. | | Architecture https://github.com/perplexityai/SPACE/blob/main/docs/architecture.md | How templates, sandboxes, processes, sessions, and operations relate; what pause and suspend preserve; how the services fit together. | | Installation and configuration https://github.com/perplexityai/SPACE/blob/main/docs/installation.md | Host prerequisites, local Kind deployment, configuration defaults, authentication, verification, and teardown. | | Operator CLI https://github.com/perplexityai/SPACE/blob/main/crates/apps/admin/cli/README.md | API access, port forwarding, space-admin-cli login , nodes, warm pools, and system-template builds. | | Your first sandbox https://github.com/perplexityai/SPACE/blob/main/docs/first-sandbox.md | A complete API workflow, command customization, output handling, inspection, and deletion. | | E2B Compatibility Wrapper https://github.com/perplexityai/SPACE/blob/main/docs/ecw.md | Configure ECW, create a compatibility API key, use the tested E2B Python SDK, and understand supported behavior and limitations. | | API reference https://github.com/perplexityai/SPACE/blob/main/docs/api/index.md | Endpoints, request and response fields, and examples. | | Troubleshooting https://github.com/perplexityai/SPACE/blob/main/docs/troubleshooting.md | Local setup, authentication, lifecycle operations, command streams, and safe retries. | | Operator egress profiles https://github.com/perplexityai/SPACE/blob/main/docs/egress-profiles.md | Configure named policy additions for federated native and ECW creates. | | Observability https://github.com/perplexityai/SPACE/blob/main/docs/observability.md | Telemetry build features, the Rust/Go selector contract, and unchanged image defaults. | | Warm pools https://github.com/perplexityai/SPACE/blob/main/docs/warm-pools.md | Keep booted sandboxes ready on Kubernetes so creates skip a cold boot. | | Base snapshots https://github.com/perplexityai/SPACE/blob/main/docs/base-snapshots.md | Restore sandboxes from a snapshot on a single node, optionally with a node warm pool. | To browse the guide and API together locally, install mdBook 0.4.52 https://github.com/rust-lang/mdBook/releases/tag/v0.4.52 and the API build prerequisites https://github.com/perplexityai/SPACE/blob/main/docs/api/index.md browse-locally . From a clean checkout, run: bash scripts/build-docs.sh python3 -m http.server 8000 --bind 127.0.0.1 --directory target/docs/site Open http://127.0.0.1:8000 http://127.0.0.1:8000 . The build also requires Git and Python 3.11+. The standalone commands remain bash scripts/build-guide-docs.sh and bash scripts/build-api-docs.sh ; they also support local uncommitted edits. The API serves its generated reference at /openapi-m1.json , with request fields, response schemas, and errors. The SPACE API manages sandbox lifecycles, runs and controls processes, and streams their output. Fields marked incubating in the schema are outside the supported contract. SPACE is implemented in Rust, but applications interact with it over HTTP and server-sent events. You do not need to embed a Rust library to use it; the first-sandbox walkthrough runs a standard-library-only Python example. Installation and cluster diagnostics use Bash. See DEPENDENCIES.md https://github.com/perplexityai/SPACE/blob/main/DEPENDENCIES.md when adding or updating third-party code. After local installation https://github.com/perplexityai/SPACE/blob/main/docs/installation.md , run tilt up from the repository root to rebuild affected images and roll their workloads as files change. The Kind runbook https://github.com/perplexityai/SPACE/blob/main/infra/space/kind/README.md covers this development loop, local topology, and template builds. Use focused unit tests for component logic and the Kind smoke suites https://github.com/perplexityai/SPACE/blob/main/infra/space/kind/README.md smoke-test for runtime behavior. For setup-specific diagnostics, see host-path troubleshooting https://github.com/perplexityai/SPACE/blob/main/infra/space/kind/README.md host-paths-and-running-as-a-user-other-than-dev and template-build debugging https://github.com/perplexityai/SPACE/blob/main/infra/space/kind/README.md debugging-a-template-build . crates/apps/ CLI tools and service binaries crates/libs/ Shared Rust libraries docs/ Architecture, setup, API walkthrough, and troubleshooting examples/ Runnable API examples without an SDK dependency proto/ gRPC protobuf definitions templates/firecracker/ Seed builder and template recipes scripts/ Utility scripts and smoke suites scripts/tests/ docker/ Service Dockerfiles and Rust artifact build ci/ CI checks and artifact build scripts infra/space/kind/ Local Kind configuration, bootstrap, and development runbook infra/space/aws/reference/ Reference EKS deployment, applied on a schedule SPACE is licensed under the Apache License, Version 2.0 https://github.com/perplexityai/SPACE/blob/main/LICENSE . See CONTRIBUTING.md https://github.com/perplexityai/SPACE/blob/main/CONTRIBUTING.md to propose changes or contribute. For native API contract, concurrency, recovery, and quarantine coverage, see Native API hardening https://github.com/perplexityai/SPACE/blob/main/docs/native-api-hardening.md .