Bot & AI
Web-based chat simulator that runs your bot end-to-end — same LLM, same tools, same knowledge base — without sending real WhatsApp messages. Catch every bug before going live.
As production
1:1 fidelity
No phone needed
Browser-only test
Inspector
See state every turn
Recording
Save + replay tests
Same LLM, same tools, same knowledge base, same flow logic as production. If it works in the simulator, it works for customers.
After every message, see exactly what variables the bot has captured. Is it remembering the user's name? Did it extract the order ID? Debug in real time.
See every tool call the LLM makes, with arguments and response. Catch hallucinated tool calls or wrong argument extraction.
Record a test conversation, save it with a name, replay anytime. Build a regression suite — re-run all scenarios after every bot change.
Export simulator sessions as JSON for debugging or sharing with teammates. Attach to bug reports; replay in different bot versions.
Clear the conversation and start fresh with one click. Useful when testing edge cases without polluting state from earlier turns.
Built a new bot or updated a system prompt? Test in the simulator first. Verify the happy path, then explicitly test edge cases.
Use scenario recording to save common conversations: 'standard order inquiry', 'angry customer escalation', 'multilingual code-mix', 'out-of-scope question'.
When a real conversation goes wrong, copy the customer's messages into the simulator. Reproduce the bug, fix the bot config, verify the fix works.
New marketing hire? Have them play with the simulator before touching production. They learn the bot's capabilities risk-free.
Every bot change should be smoke-tested against your saved scenarios. The simulator runs them quickly and tells you what broke.
| Feature | Edesy | Typical platform |
|---|---|---|
| Web-based bot simulator | Send-to-self only | |
| Same engine as production | Approximation | |
| Variable state inspector | ||
| Tool call visibility | ||
| Scenario save + replay | ||
| Export conversations | ||
| Voice message testing | ||
| Multi-version testing |
QA team runs 30 saved scenarios against the new bot before publishing. Catches 80% of bugs before customers see them.
Production bug rate down from 4/week to 0.5/week
Marketing manager iterates on the system prompt. Each change tested against same scenarios in the simulator before publishing.
Bot improvement cycle: 1 day vs 1 week previously
Support agent gets a complaint that the bot misunderstood. Copies the conversation into the simulator, sees the issue, files a fix.
Most reported bot issues fixed within 24 hours
Test scenarios in English, Hindi, Spanish, Tamil. Verify the bot responds correctly in each language.
Multilingual rollouts ship without language-specific bugs
After connecting a new custom tool (e.g. Shopify order lookup), test in simulator with realistic queries before exposing to customers.
Tool integrations ship reliably, no 'API broke at launch' incidents
Sales rep rehearses the bot demo for upcoming customer call. Runs through the exact conversation they'll show, fixes any rough spots.
Sales demos always go smoothly
The state of WhatsApp bot development is suspiciously similar to web development circa 2003: you make a change, you wait, you 'test in production', and you hope. Every other platform's 'test' button just sends a real WhatsApp message to your own number — useful for the most basic smoke test, useless for systematic QA. You can't easily test edge cases, can't replay scenarios, can't inspect state at each turn, can't debug failures.
Edesy's bot simulator is built to give bot developers the same iteration experience modern web devs take for granted. It runs the same exact engine as production — same LLM provider, same model, same temperature, same tools, same knowledge base, same flow logic. If it behaves a certain way in the simulator, it behaves the same way for customers. There's no 'production drift' where the simulator passes but real conversations fail.
The variable state inspector is the feature most teams underestimate before they use it. After every customer message, you can see exactly which variables the bot has captured — is it remembering the user's name? Did it extract the order ID into the right field? Is the context window getting bloated with old turns? — and which tools it has called. Most bot bugs are invisible from the outside but obvious from this inspector. It cuts debugging time by 5-10x.
Scenario save + replay is the feature that elevates the simulator from 'nice utility' to 'core to your dev workflow'. Save 20-30 representative conversations as scenarios — happy path, edge cases, languages, complaints, escalations. After any bot change, replay all scenarios in 5 minutes and see what broke. This is regression testing for bots, and it transforms the change-and-pray bot development cycle into a fast, confident iteration loop.
Free workspace, full simulator from day one. Most teams ship their first bot improvement within hours of opening the simulator.