Back to Parloa brief

AgentStatus × Parloa · Integration

TextchatV2 sandbox. We are ready to wire.

Public OpenAPI is enough for day one. Share a staging release and we run outside-in probes beside your Test stack.

Prepared from public docs

We studied TextchatV2 before this conversation

The pilot path is clear: a sandbox (or staging) textchat release, then outside-in probes against dialoghook. We do not need to scrape customer sites or dial production voice lines to start.

Deep AMP product docs sit behind login (docs.amp.parloa.com). What is public is enough to wire the moment you share credentials. Whitelist on AMP docs is welcome; it is not a blocker for day one. Source: parloa.github.io/chat-api-spec.

What we need from you

Four items to start the pilot

01

Host

Confirm stage.parloa.com or app.parloa.com for the sandbox release. Your public OpenAPI lists both.

02

Release credentials

releaseId (query param dialog) and apiToken (Authorization: Bearer), minted when the textchat release is created. Plus expiry or rotate process if any.

03

Guardrails

Allowed probe rate, concurrent sessions, and confirmation that the sandbox holds no production customer PII.

04

Scenarios

Three to five scripts you want scored, plus known happy paths and known fail cases. FAQ-only is fine for week one; tool or CRM mocks if the sandbox has them.

How we will call it

dialoghook protocol

Single endpoint, request-response, per your public spec:

POST {base}/dialoghook?dialog={releaseId}&platform=textchatV2
Authorization: Bearer {apiToken}

1. event.type = launch   → welcome
2. event.type = message  → turn text (same sessionId)
3. event.type = quit     → or end when endConversation=true

We keep sessionId stable for multi-turn state, use a distinct requesterId per probe identity, and score reachability, empty or broken replies, conversation survival, and latency. Optional context key/values are supported if your sandbox expects them. We will report conversationId / request ids back when something fails so your team can look it up.

Where this sits

Beside Test, voice, and CRM

Three layers, one pilot surface

TextchatV2 is the integrator path into a text agent. That is our pilot surface.

Voice and CCaaS (Genesys, Five9, NICE, and similar) is how most enterprise agents go live, often SIP plus transfer context. We are not asking to onboard that for a first pilot. Voice outside-in is roadmap for us; we say that plainly.

CRM and backend skills (Salesforce, ServiceNow, SAP, and the rest) live inside the agent you already built. We do not integrate those systems. We exercise the deployed agent that already has them.

Your Test stack (sim, judges, post-deploy transcript eval) stays the builder-stack system of record. We sit beside it as residential outside-in on the same sandbox agent, not as a second sim engine.

Closing

Ready when the release credentials are.

Share a TextchatV2 sandbox and we run the outside-in layer beside Test. You keep the report either way.

Contact·dulra@carmel.so·roman@carmel.so

Based on Parloa's public TextchatV2 OpenAPI (parloa.github.io/chat-api-spec). AMP product docs behind login were not used as a source. Protocol details may differ slightly by tenant; we will confirm host and release fields with your team before probing.