Your first sandbox check.
No account, API key or personal data required. Python 3.11+.
Research and three local utilities run offline after installation. The five merchant profiles send authorized task data to your configured GridZen API over HTTPS and require an existing merchant key, current permissions and the accepted quote. They do not inherit the anonymous research service’s no-provider behavior. Installing a profile does not activate paid access.
1. Download and install
Download the developer kit, unzip it, and run from its parent directory:
python3 -m venv .venv . .venv/bin/activate python -m pip install --upgrade gridzen-developer-kit gridzen setup --client all --project . gridzen setup --client all --project . --apply gridzen plan --country ID --event payout
On Windows, activate with .venv\Scripts\activate. CLI and local MCP work offline after dependency installation.
2. Test the failure cases
gridzen simulate --country ID --capability bank_account_match --scenario timeout gridzen simulate --country MX --capability bank_account_match --scenario not_found
The Python example calls the public sandbox and tests all five outcomes. A timeout or missing record is inconclusive. All results remain simulated=true and verified=false.
3. Connect your Agent
Connect directly to the public research/sandbox MCP endpoint. No account or API key is required:
Choose your client and see its configuration →
gridzen setup --client codex --project . gridzen setup --client codex --project . --apply
Ask: “Plan a payout verification integration for Indonesia, then test a provider timeout.” Five tools cover research, planning, simulation, fixture retrieval and explanation. Remote requests go to Gridzen; local stdio remains an offline option. Client configuration formats vary.
For offline use after installation, register the absolute path to gridzen-mcp as a local stdio command.
4. Install a Skill
Install the fourteen Skills from the public repository:
npx skills add immurray/gridzen-developer-kit
Download the fourteen Skills and copy their individual folders into your agent's supported skill directory. Keep each SKILL.md and its references together. Select verification, integrate a sandbox, or explain a result.
HTTP API
curl https://gridzen.ai/developers/api/plan \
-H 'Content-Type: application/json' \
-d '{"country":"ID","event":"payout"}'
curl https://gridzen.ai/developers/api/sandbox/verifications \
-H 'Content-Type: application/json' \
-d '{"country":"ID","capability":"bank_account_match","scenario":"not_found"}'Download the OpenAPI schema. Requests accept only documented country/event/capability/scenario fields. No personal data or provider credentials are accepted. Synthetic IDs identify reusable fixtures, not real transactions.
Connect it to an app
The kit includes a Python HTTP SDK and a Next.js server-route example. Use a bounded network timeout, validate inputs, keep an explicit sandbox indicator, and handle unknown or unavailable responses separately from mismatches.
Ready for a real pilot?
Research is dated 2026-10-07. Sources and access requirements are included; enabled live routes remain zero. Contact open@gridzen.ai with the country, capability and expected volume to evaluate an authorized route.
Choose the customer task
gridzen plan --country US --event onboarding --task signup_phone gridzen plan --country US --event onboarding --task identity_onboarding python examples/customer_tasks.py
Phone type does not verify ownership or send OTP. Identity fixtures do not create hosted document sessions. Production plans return blockers. Optional local feedback: gridzen feedback --task signup_phone --country US --outcome blocked --blocker missing_workflow_step. Upload requires an explicit --share.