Playground Onboarding
New self-serve workspaces land in the Playground at the workspace home. Its centerpiece is the "Ask your documents" card: a question box wired to the agent, sitting over a workspace that is already seeded with two example documents, a Northwind Supplies invoice and an Acme/Globex master services agreement, both fully ingested with captured fields. Nothing needs to be configured or uploaded before the first question: type "What is the total value of the master services agreement?" and the first cited answer arrives in seconds.
The seeded examples exist so the first thing you experience is the loop that matters: ask, read, verify. Each answer cites its sources inline, and clicking a citation opens the document with the cited span highlighted, so you can see for yourself that the "EUR 96,000" in the answer is the "EUR 96,000" in the contract. The verification badge on the answer shows whether a second pass could support every claim. Once that loop feels trustworthy on the examples, it works identically on your own documents.
From seeded examples to your own documents
The Playground flow for real data is upload, then ask. Drop your own files onto the Playground (or ingest them programmatically) and they run through the same pipeline as the examples: OCR, then background field capture that adds them to the askable corpus. As soon as processing completes, questions span everything in the workspace, seeded examples and your uploads alike, and you can delete the examples once they have served their purpose. The developer path mirrors this: mint an API key and the same two calls, POST /v1/run to ingest and POST /v1/ask to question, reproduce the whole Playground experience over HTTPS.
# 1. Ingest your document (askable by default)
curl -s -X POST https://api.talonic.com/v1/run \
-H "Authorization: Bearer $TALONIC_API_KEY" \
-F "spec_id=$SPEC_ID" \
-F "files=@my-contract.pdf"
# 2. Ask about it
curl -s -X POST https://api.talonic.com/v1/ask \
-H "Authorization: Bearer $TALONIC_API_KEY" \
-H "Content-Type: application/json" \
-d '{"question": "What are the payment terms in my contract?"}'For developers who signed up through the magic-link flow (on deployments where self-serve signup is enabled), the Playground and the API key arrive together: registering by email confirms into a seeded workspace plus a scoped tlnc_ key, so the first curl works within a minute of signup, against the same Northwind and Acme/Globex examples the Playground shows.
What the Playground sidebar shows
A Playground workspace has a deliberately short sidebar with five entries: Home (the Ask your documents card and your uploads), API Keys (mint and revoke keys for the developer path), Usage (what your questions and uploads have consumed), Billing, and Account. The full platform sections, such as Documents, Registry, Spec, Pipelines, Products and Apps, do not appear until the workspace has platform access. The Agent (Cmd+I) and search (Cmd+K) work the same way in both modes, so the habits you form in the Playground carry over directly.
Because the Playground and the platform share one corpus, the work you do in one is visible in the other. A document uploaded on the Playground home is the same document that later appears on the Documents page, and a concept an ask promoted into the structured layer is the same concept a Spec can later bind to in the Field Registry. Nothing has to be re-ingested or re-extracted when the workspace graduates, and every citation link you saved keeps resolving to the same document and span.
Graduating from the Playground
The Playground is the full ask experience, but it is deliberately a subset of the platform: self-serve workspaces start with the ask surface, uploads, and the API, while the complete platform workspace (Specs and pipeline composition, review on Products, delivery, team administration) is part of the enterprise plan. When your usage outgrows the Playground, use the request-access path in the app to talk to the team about upgrading; your workspace, documents, captured fields, and API keys carry over unchanged, because the Playground runs on the same corpus and the same engine as the full platform.
Until then, nothing about the Playground is throwaway. Every question you ask makes the corpus more structured (promotion and mining persist their results), every upload is a real ingested document with provenance, and credit metering works the same as everywhere else: one flat agent_ask unit per question. Teams routinely run the Playground as their production ask surface over the API long before touching the rest of the platform.