Already running a bot on another platform? You don't have to rebuild it. Convoship imports your bot's export, recognizes the format automatically, and converts the intents, entities, and conversation flow into Convoship's own schema — ready to edit on the canvas.
The same import can become either a deterministic, scripted Assistant (you control every node) or an autonomous AI Agent (mission, tools, persona, and guardrails). You choose at the last step.
What you can import
Export your existing bot from your current platform — usually a .zip archive with its intents, entities, and conversation flows — and drop it in. You never pick the platform by hand: Convoship inspects the contents of the archive and auto-detects the format.
Auto-detected
You don't pick the platform — Convoship inspects the contents of the export archive and detects it for you. If the file isn't a recognized bot export, you'll get a clear message rather than a wrong guess.
How it works
- Upload — open Import on the Assistants page (or the 'Import a legacy bot' card) and drop in your export .zip.
- Detect & map — Convoship detects the platform and shows what it extracted: intents, entities, flow nodes, and any tools it mapped from your service calls.
- Ship — choose your output: a deterministic scripted Assistant, or an autonomous AI Agent. Convoship creates it and takes you straight to the canvas.
Choose your output
| Output | Best when | What you get |
|---|---|---|
| Deterministic Assistant | You want the same scripted control you had before | A node-by-node flow you edit on the canvas, with versioning, live simulation, and the drop-off heatmap. |
| AI Agent | You want the model to reason over the long tail | An agentic runtime with a mission, tools, persona, and guardrails — plus evals and spend caps. |
Not sure which?
Start with a deterministic Assistant for predictable, regulated flows, and upgrade it to an AI Agent later in one click. See Agent vs Assistant for the full comparison.
What gets converted
- Intents and their training phrases.
- Entities and slots, including system types and custom lists.
- Conversation flow — pages, states, and the transitions between them.
- Service / fulfillment calls, lifted into Convoship tool nodes you can wire to your APIs.
Works offline
Enrichment uses a single model pass to clean labels and infer slots, with a built-in heuristic fallback. If no model provider is configured, the import still completes using the deterministic baseline — it never silently fails.
After you import
Your migrated bot lands as a normal Convoship assistant or agent. From there you can edit it on the canvas, simulate it live, give an AI Agent cross-session memory, watch the per-node drop-off heatmap once it's taking traffic, and publish to any channel. Nothing about the import is a dead end — it's the same workspace as a bot you sketched from scratch.
Keep improving it
Once live, the drop-off heatmap shows exactly where conversations stall, and the flow doctor and persona simulation help you fix weak spots before customers hit them.