Harness testing
Give every test run an isolated sandbox. Changes persist across tool calls, then reset to the same baseline.
Testing sandboxes, built for AI agents
Test API integrations without test accounts, rate limits, or dirty state. Each sandbox contains stateful environments that behave like the APIs you use in production.
Quickstart ↓Get $20 in usage credits
Backed byCombinator
Choose environments for your sandbox:
ApifyCal.comClickUpCustomer.ioDatadogDiscordGitHubGitLabGoogle WorkspaceHubSpotJiraLinearOwnerRezPriceLabsRampSentrySlackStripeSupabaseTavilyUnipileWebflowApifyCal.comClickUpCustomer.ioDatadogDiscordGitHubGitLabGoogle WorkspaceHubSpotJiraLinearOwnerRezPriceLabsRampSentrySlackStripeSupabaseTavilyUnipileWebflow
Use cases
Built for agent developers, connectors teams, and integration engineers.
Integration testing
Give every branch or CI job its own sandbox without shared test accounts, rate limits, or dirty state.
Agent training and RL
Give every rollout a sandbox with the same starting state while changes persist through the episode.
Product →
Each sandbox holds one or more stateful API environments. Changes persist between calls and reset to the same starting point.
Create, call, inspect, reset
+
Create
Create a sandbox with one or more environments and a known starting state.
✓
Ready
Receive isolated REST and MCP endpoints as each environment starts.
→
Call
Use familiar API paths, payloads, and tools.
⌁
Inspect
See requests, responses, mutations, timing, and usage together.
↻
Reset
Restore every environment in the sandbox to its declared baseline.
Choose the environments your sandbox needs
Every environment is stateful and resets to a declared starting state.
GitHubRepos, issues, pull requests, branches, and commits.LinearIssues, projects, teams, cycles, and workflows.SlackChannels, messages, threads, users, and reactions.SupabaseSQL, migrations, logs, branches, and project metadata.ApifyActors, runs, datasets, key-value stores, and request queues.Cal.comEvent types, schedules, availability slots, bookings, calendars, and webhooks.ClickUpTeams, spaces, folders, lists, tasks, comments, tags, time tracking, and checklists.Customer.ioMessaging automation across campaigns, broadcasts, transactional email, segments, customers, and message delivery.
View environments →
Each environment is an independently built simulation of the service it represents, not the provider’s live service.
Let your coding agents set you up
Paste the skill below into Claude Code, Codex, Cursor, or Devin. It starts secure signup, installs five verified skills, and checks access to every environment.
Get $20 in free usage
No card required. $5 arrives upon sign up and the remaining $15 when your first sandbox is started.
- Verified signup
- $5
- First sandbox started
- +$15
For your coding agent
Set up Archal in this repository.
1. Detect whether I am using Claude Code, Codex, Cursor, or Devin.
2. Use target `portable` for Devin; it installs the verified Agent Skills bundle in `.agents/skills`. Otherwise use the matching target.
3. Run `npx --yes archal@0.11.3 connect --device --target <target> --non-interactive`.
4. Show me the verification URL and one-time code, then wait while I create or approve my account in the browser.
5. Do not continue until the command verifies all five skills and access to the complete Archal environment catalog.
Do not modify application code or start an environment during setup. Only after the command exits successfully, reply exactly: Archal is connected and readyFrequently asked questions
An Archal sandbox is an isolated, time-bounded runtime containing one or more stateful environments. Each environment exposes familiar REST or MCP interfaces, retains mutations for the life of the sandbox, and can return to its declared starting state.
No. One Archal API key creates and manages every sandbox and environment. Calls share one gateway while each environment keeps the familiar paths, methods, payloads, and response shapes your existing code expects.
State behavior varies by environment and operation. Available environments have a supported capability profile; environments in progress remain callable only where their documented behavior permits it. We do not claim complete upstream coverage without measured, dated evidence.
Yes. Choose every environment your workflow needs when you create the sandbox. Each environment receives its own endpoint and readiness state. Each environment costs $0.10 per minute.
Environment traffic is routed to Archal-managed implementations, not your production accounts. Each environment profile documents any exceptional dependency or behavior before you use it.
Self-serve usage has no subscription and requires no card to start. New accounts receive $5 after verified signup and another $15 when their first sandbox is started. Each environment costs $0.10 per minute. Billing is prorated by the second; failed provisioning and cold-start time are free.