Home

Twin environments

See the environment your agent runs in.

Every twin you create — from your own codebase, from the service catalog, or as a blank test harness — appears here as a live environment graph: gateway, service twins, runtime, verified state, sessions, and a digest-only activity ledger. Sign in to create and view yours.

Sign in to continue

Using an API key instead? The same data is on GET /api/v1/twin/environments and GET /api/v1/twin/runs with your X-API-Key.

Who it’s for

Frontier labs · neolabs · teams training their own agents

RL training environments

Blobfish gives models multi-app workflows where they can practice real tool use, acquire skills, and improve before touching production.

Skill acquisition needs real environments. We believe intelligence should be measured by how efficiently a model can acquire new skills, not just by raw performance on a fixed task. Current LLMs can look impressive on internal benchmarks, but they are weak at acquiring new operational skills in open-ended software environments — repeatable multi-app workflows are where models learn to handle tools, state, permissions, failures, and long-horizon coordination.

The hard part is using external tools. Most useful systems will not live inside a single internal tool. They will use browsers, APIs, databases, SaaS products, internal systems, and workflows that change over time. Future models should not just know how to call tools; they should learn how to learn tools — try, observe, adapt, and transfer what they learn to new software.

Enterprise

Enterprise agent sandboxes

Run agents against realistic twins of services like Slack, Stripe, GitHub, and Jira, with ground truth of good tool calls. Production-shaped testing, without production consequences.

  1. Ground-truth tool calls for each workflow
  2. Resettable twins agents can train against safely
  3. Inspectable state, traces, and outcomes for evaluation
Startups

Code-change validation

Production-shaped testing, without production consequences.

  1. Deterministic browser runner that clicks through your app
  2. Validation against third-party integrations and API twins
  3. Reports with traces, screenshots, side effects, and failure context

Service twins

Deterministic, resettable twins of production services — seeded state, REST + MCP + CLI surfaces, and per-session isolation. 28 of 28 available twins are mounted on this deployment; each card says exactly which runtime backs it.

Attio

captured contract

Test CRM agents that manage people, companies, relationships, and sales workflows.

API + UI  34 operations · 26 entities

Box

captured contract

Test enterprise content agents that organize files, manage collaboration, and control sharing.

API only  273 operations · 191 entities

Check

docs-transcribed

Test payroll agents and integrations that onboard workers, run payroll, and coordinate payments and filings.

API + UI  31 operations · 7 entities

/api/mcp/check

Discord

captured contract

Test community and support agents that moderate servers, answer users, and coordinate channel workflows.

API + UI  128 operations · 71 entities

Documenso

captured contract

Test document agents that prepare envelopes, route recipients, collect signatures, and manage signing workflows.

API + UI  17 operations · 13 entities

Dropbox

captured contract

Test file agents that organize content, manage sharing, and retrieve documents across team folders.

API + UI  12 operations · 9 entities

GitHub

spec-grounded

Test coding agents that manage issues, review pull requests, and coordinate repository workflows.

API + UI  1,108 operations · 400 entities

/api/mcp/github

GitLab

spec-grounded

Test coding and DevOps agents that manage projects, merge requests, issues, and CI/CD pipelines.

API + UI  1,108 operations · 421 entities

/api/mcp/gitlab

Gmail

spec-grounded

Test email agents that triage inboxes, draft replies, and manage message workflows.

API + UI  79 operations · 41 entities

/api/mcp/gmail

Google Calendar

spec-grounded

Test scheduling agents that create events, resolve conflicts, and coordinate attendees.

API + UI  38 operations · 11 entities

/api/mcp/googlecalendar

Google Docs

spec-grounded

Test document agents through the Google Docs API and an interactive rich-text editor.

API + UI  4 operations · 3 entities

/api/mcp/googledocs

Google Drive

spec-grounded

Test document agents that find, organize, edit, and share files across team drives.

API + UI  64 operations · 40 entities

/api/mcp/googledrive

Google Sheets

spec-grounded

Test spreadsheet agents through the Google Sheets API and an interactive grid editor.

API + UI  18 operations · 16 entities

/api/mcp/googlesheets

Google Workspace

docs-transcribed

Test agents across Google People, Workspace Events, Apps Script, and Pub/Sub without calling production APIs.

API only  34 operations · 7 entities

/api/mcp/googleworkspace

HubSpot

spec-grounded

Test sales and support agents that update CRM records, move deals, and manage customer workflows.

API only  28 operations · 7 entities

/api/mcp/hubspot

Jira

captured contract

Test engineering agents that triage issues, update sprints, and coordinate delivery workflows.

API only  46 operations · 35 entities

Linear

captured contract

Test engineering agents that triage issues, plan projects, and coordinate product work.

API only  25 operations · 12 entities

LinkedIn

docs-transcribed

Test professional-network agents that research people and companies and manage content workflows.

API + UI  29 operations · 6 entities

/api/mcp/linkedin

Notion

spec-grounded

Test knowledge agents that search, update, and organize team documentation.

API + UI  24 operations · 14 entities

/api/mcp/notion

Outlook

Roadmap

Planned twin for Microsoft mail and calendar workflows across messages, folders, events, and attachments.

API + UI

QuickBooks

captured contract

Test finance agents that manage customers, invoices, bills, payments, purchases, and accounting records.

API + UI  18 operations · 11 entities

Resend

captured contract

Test email agents and applications that send, receive, schedule, and track transactional messages.

API + UI  65 operations · 22 entities

Salesforce

domain twin

Test revenue agents that manage leads, opportunities, accounts, and customer follow-up.

API only  52 operations · 9 entities

/api/mcp/salesforce

Slack

spec-grounded

Test workplace agents that answer questions, coordinate teams, and act across channels and threads.

API + UI  174 operations · 61 entities

/api/mcp/slack

Stripe

spec-grounded

Test commerce agents that manage subscriptions, payments, invoices, and customer billing.

API + UI  587 operations · 203 entities

/api/mcp/stripe

Trolley

docs-transcribed

Test payout agents that onboard recipients, manage compliance, and coordinate global payment workflows.

API + UI  28 operations · 5 entities

/api/mcp/trolley

Twilio

Roadmap

Planned twin for communications agents that handle messages, calls, conversations, phone numbers, webhooks, and delivery events.

API only

Unified

docs-transcribed

Test integration agents built against Unified.to across multiple connected providers.

API only  28 operations · 6 entities

/api/mcp/unified

Unstructured

docs-transcribed

Test document-processing agents that partition files and run ingestion workflows.

API only  26 operations · 6 entities

/api/mcp/unstructured

Waterfall

domain twin

Test revenue and GTM agents that find prospects, enrich contacts and companies, and verify work emails.

API only  25 operations · 5 entities

/api/mcp/waterfall

Plus 7 domain twins (banking, healthcare, law-firm, logistics, manufacturing, hospitality, records-governance) and 5,979 registry-captured MCP servers in the full mock catalog.