Skip to content

Differently documentation

Start with the project README for what Differently is and a two-minute local quickstart.

Guide For
Quickstart Laptop dev mode, first sandbox, SDK snippets
Organization onboarding One installation per organization vs shared; orgs, projects, quotas, scoped keys, policies
Hybrid overflow Authorizing Daytona or Freestyle as policy-gated external capacity
OpenWork on Differently OpenWork’s sandbox contract mapped onto Differently; Workbot computers routed across OpenSandbox, Daytona and Freestyle
Keys and providers Storing provider keys in Differently, switching providers for new sandboxes, rotating provider and API keys, the bootstrap token, backing up the secrets key
OpenSandbox pools The default private runtime: registering an OpenSandbox server you (or someone else) run
Admin console The separate admin app: static files on any host, or on Node with Better Auth sign-in; CORS and reaching a private Differently
Provider catalog Every backend Differently can place sandboxes on, with its tested status, capabilities and configuration, and how to add one
API reference OpenAPI 3.1 contract (also served at /openapi.yaml)
SDKs TypeScript, Python, Go
Guide For
Deployment overview and support matrix Where to run it, and what is tested, documented or unverified
One Linux VM on any cloud The quickest team setup: k3s, OpenSandbox and Differently on Azure, AWS, Google Cloud or bare metal
Prerequisites Cluster, Postgres and an OpenSandbox server (or another backend)
Kubernetes (existing cluster) Helm install to first sandbox
AWS · GCP · Azure · Bare metal Environment-specific notes
Air-gapped Mirroring artifacts, no-egress operation
Day-2 operations Upgrades, migrations, backups, drains, uninstall
Hardening checklist Before production
Document Contents
Feature parity Differently vs Boat, Daytona, Freestyle, E2B, OpenSandbox; backend capability matrix
Competitive landscape Ownership models and where Differently fits
OpenSandbox spike What OpenSandbox offers as a backend or foundation, observed and read
Architecture RFC Layers, components, request lifecycle, state machine, HA
Security and trust model Threats, boundaries, enforced controls, residual risks
Adapter specification The backend contract and the HTTP plugin protocol
Maintaining and improving Differently For agents and maintainers: where to change what, the verification ladder, recorded runtimes, performance, traps
Placement policy Merge rules, strategies, overflow, quotas, fairness
Storage and lifecycle Persistence, snapshots, fork, portability limits
Reliability and performance SLOs, benchmarks to run, chaos tests
Acceptance scenarios Scenarios A–L and the tests that cover them
End-to-end suite The real stack against an OpenSandbox server, and where it was verified
Document Contents
Roadmap Phases 0–4 with current status
Backlog Epics and tasks with acceptance criteria
Proofs of concept Risks and the experiments that retire them

Source: docs/README.md