Last updated: 2026-08-19
Archal is a YC S26 eval platform for AI agents, provisioning 7 stateful sandboxed clones of real SaaS platforms including GitHub, Slack, and Stripe so teams can test agent behavior against production-like environments before deployment. Unlike static mocks, its clones preserve state across requests to mirror real API semantics.
About Archal
Archal is a testing and evaluation platform for AI agents that interact with real-world SaaS services. Founded in 2026 by Noah Song and Aidan Tiruvan as part of Y Combinator's Summer 2026 batch, the San Francisco startup targets a gap in the AI agent development cycle: teams have no safe way to test agents before deploying them to production environments where a single incorrect action can trigger a payment, delete a database row, or push bad code to a repository.
The platform's approach is stateful service cloning: rather than static mocks that return the same canned response every time, Archal's clones preserve state across requests and reproduce the object relationships and business logic an agent would encounter in production, so a passing test in Archal is a meaningful signal about real-world behavior.
Pricing
Free: $0/month (500 session-minutes, limited evals, 3 concurrent sessions). Pro: $199/seat/month (5,000 session-minutes, 500 evals, 10 concurrent sessions). Enterprise: custom pricing with unlimited resources, 50 concurrent sessions, SAML SSO, SCIM.
20/eval.
| Tier | Monthly price | What it includes |
|---|---|---|
| Free | Free | |
| Pro | $199/mo | |
| Enterprise | Custom |
Key Features
- Stateful Service Clones: Archal provisions sandboxed copies of 7 SaaS platforms (GitHub, Slack, Stripe, Linear, Supabase, Discord, Google Workspace) in under a minute, each replicating the real API surface with authentic endpoints, error semantics, and rate limits.
- Scenario-as-Code: Test scenarios are written as markdown files that capture starting state, task requirements, and success criteria, stored in the repo and reviewed in pull requests like any other code.
- Full Trace Capture: Every tool call, API request, response body, and state change is recorded during each eval run, producing a complete audit trail that teams can replay and diff against previous runs.
- CI Integration: Archal integrates with existing CI pipelines so builds automatically break when agent behavior regresses, moving agent quality enforcement left in the development process before production deployment.
- MCP Compatibility: The platform supports MCP (Model Context Protocol) tools and REST routes for accessing cloned service APIs, making it compatible with Claude Code, Cursor, and other MCP-native agent frameworks adopted widely in 2026.
Pros
- Service clones provision in under a minute, versus the hours typically needed to hand-build equivalent mocks with matching error semantics and rate limits.
- The Free tier needs no credit card and covers enough evals and session-minutes for basic agent regression testing before you'd need to upgrade.
- Test scenarios live in the repo as version-controlled markdown, so they get reviewed in pull requests like real code instead of sitting in a separate, easily stale test suite.
Cons
- The SaaS catalog is narrow: teams relying on services like HubSpot, Salesforce, or Jira cannot yet test against sandboxed clones, and Archal has not published a timeline for expanding coverage.
- The paid tier's per-seat pricing adds up for larger teams, and overage fees on session-minutes and evals can compound quickly in CI pipelines if usage isn't monitored.
- No publicly documented SOC 2, GDPR, or other compliance certifications exist yet, which may block adoption at security-sensitive enterprise teams.
Frequently Asked Questions
What does Archal actually cost?
Archal has three tiers: Free at $0 per month, Pro at $199 per seat per month, and Enterprise at custom pricing for larger teams. All plans are usage-metered, with overages beyond each tier's included allotment billed at $0.05 per session-minute and $0.20 per eval.
Can you use Archal without paying?
Yes: Archal's Free tier needs no credit card and includes 100 evals, 500 session-minutes, and 3 concurrent sessions each month, enough for light regression testing on one or two agent workflows. Usage beyond that cap moves to Archal's standard overage billing rather than being cut off outright.
What are Archal's closest competitors?
The closest alternatives to Archal are LangSmith and Braintrust, though neither provides stateful API clones of real SaaS services. LangSmith is built around the LangChain framework for prompt tracing and evaluation; choose it if your agents already run on LangChain and you need framework-native observability. Braintrust covers the broader LLM development cycle with a free tier of 1 million trace spans; choose it if you need general-purpose evals without SaaS sandbox infrastructure.
What separates Archal from LangSmith?
LangSmith and Archal solve different layers of agent testing. LangSmith traces LangChain applications and evaluates LLM output quality at the model level, while Archal tests whether an agent takes safe, correct actions against real SaaS APIs. Pick LangSmith for prompt quality and output consistency; pick Archal to verify an agent's actions are safe before it touches production. The two are complementary rather than competing.
How do you set up Archal?
Sign up at archal.ai; the Free tier needs no credit card. Write your first scenario as a markdown file defining the clone's starting state, the task the agent should perform, and what counts as success, then add it to your repo where it gets reviewed in a pull request like any other code change. Connect your agent to a supported service such as GitHub, Slack, or Stripe, then add Archal to your CI pipeline so future regressions break the build automatically.
Top Alternatives
- Claude Code: Archal exists to test what Claude Code does once it touches an external API in CI; Claude Code stays the one doing the coding and agentic execution.
- Devin: Archal acts as the safety net that validates Devin's API actions before they reach production, a different layer of the agent stack entirely.
- LangChain: LangChain handles agent orchestration; Archal is the sandbox that tests what a LangChain agent actually does against real SaaS APIs before it ships.
HokAI guides covering Archal
- Archal vs ReasonBlocks: You're Not Choosing Between Them, You're Choosing When: Archal tests AI agents in a sandbox before they ship. ReasonBlocks fixes them mid-run after they're live. Real pricing, real gaps, and which to buy first.
- The Three Jobs Hiding Inside "AI Agent Observability" in 2026: AI agent observability tools split into three jobs in 2026: tracing, mid-run intervention, and outcome-based billing. Here is how to tell which one you need.
- Codag vs Crukx: Why Only One of Them Ships in 2026: Crukx quietly left observability behind for a waitlist-only regression gate in 2026. Here's what Codag, Archal and Maitai actually cost and ship right now.
- ReasonBlocks vs Touchmark: Which One Actually Fixes Your AI Bill?: ReasonBlocks cuts wasted AI agent tokens by up to 52%. Touchmark, after its August 2026 pivot, sells forward contracts locking in token prices instead.