Product documentation

Introduction

What Convoship is and how the platform fits together.

Convoship is a conversational AI studio for designing, testing, and deploying bots. Use Assistants for scripted intent + flow bots, Agents for LLM reasoning over tools, and Workflows to chain them together.

What you build

An Assistant is a scripted bot: intents, sub-flows, entities, tools, and Channels deployment tokens (cdp_*). An AI Agent is a separate agentic runtime with mission, tasks, tool handlers, publish/publicToken embed, and guardrails — see /docs/agent-vs-assistant and /docs/agentic-ai.

Studio layout

  • Workspace — top-level container for Assistants, Agents, members, secrets, and audit logs.
  • Assistants dashboard (/app/assistants) — scripted bots with intents and flows.
  • Agents dashboard (/app/agents) — agentic bots with tasks, tools, and playground.
  • Assistant editor — left rail with Build, Operate, and Move sections while editing one scripted Assistant.
  • Simulator — floating chat on every Assistant page to test conversations live.
Convoship Studio workspace showing the Assistants dashboard and main navigation
Start from the Studio workspace. The left rail separates workspace setup, scripted Assistants, agentic Agents, and Workflows so implementation work has a clear home.

Typical workflow

  1. Create a scripted Assistant from a blank canvas or imported Draw.io / PDF / Word / image source.
  2. Define intents and build sub-flows on the canvas.
  3. Add entities and tools as needed.
  4. Test utterances and simulate full sessions.
  5. Publish a version and deploy via Channels (web widget).

Implementation path

  1. Choose the bot type: scripted Assistant for deterministic journeys, AI Agent for LLM reasoning over tools, or Workflow when you need orchestration.
  2. Create the draft in Studio and keep all names user-readable because node, intent, task, and tool names appear in traces.
  3. Wire data dependencies before publish: entities, workspace secrets, HTTP tools, knowledge sources, and channel origins.
  4. Test in the Studio simulator or Playground, then inspect History, trace output, and eval results before exposing the bot publicly.
  5. Publish a version, deploy through Channels or publicToken embed, and monitor Analytics plus Change log or Audit log after launch.

Feature map

Every major studio surface has a dedicated doc page. Use the left sidebar to jump directly, or follow the table below.

Studio areaDoc page
Workspace dashboardWorkspace
Members & rolesMembers
Secrets vaultSecrets
Workspace audit logWorkspace audit log
Shared HTTP/agent toolsWorkspace tools
Plan, MFA, sessionsWorkspace settings
Assistants dashboardAssistants dashboard
New assistant wizard (3 steps)New assistant wizard
Import review gateImport review
Multi-source importImport sources
Assistant settingsAssistant settings
Flow canvas & node repairFlow builder
Intents & sub-flowsIntents
EntitiesEntities
Tools (assistant scope)Tools
Small talkSmall talk
Welcome / fallback / no-inputEvents
Knowledge (RAG)Knowledge
Python script nodesPython scripts
Runtime & intent matchingRuntime engine
Simulator & utterance testTest & simulator
Publish releasePublish
Version history & restoreVersions
Metrics & funnelsAnalytics
Web widget & providersChannels & deploy
Embed SDK APIEmbed SDK
Session transcriptsHistory
Design-time auditChange log
Agent vs Assistant choiceAgent vs Assistant
AI Agents dashboardAgentic AI overview
New Agent wizard (4 steps)New Agent wizard
Playground, convos, evalsPlayground & evals
Workflows landingWorkflows overview
Chain editorChain editor
Workflow flow canvasWorkflow canvas
JSON / Draw.io exportExport
REST API surfaceAPI reference (Developer)
Embed SDK APIEmbed SDK (Developer)

Tip

Use Change log to see who edited intents and flow. Use Versions to export or restore a past release.