🎙️✈️

Agentic AI in Action

Voice-Enabled Multi-Step Travel Booking

One conversation. Five connectors. Zero code. Full security.

The User Prompt

"Book me a flight to Denver next Friday, cheapest direct flight"

One sentence triggers an 8-step agentic workflow across 5 enterprise connectors — with conflict detection, budget validation, secure payment, and calendar updates.

What ContextWeaver Does

1
Calendar Check Microsoft Graph
check_calendar(date="next Friday")
"You have a 2pm meeting with Sarah on Friday. Should I cancel it, or try a different date?"
2
User Decides Voice Input
"Try Saturday instead"
check_calendar(date="next Saturday") → Saturday is clear
"Saturday is open. Let me find flights."
3
Flight Search Travel API
search_flights(to="DEN", date="Saturday", pref="cheapest direct")
"Found 3 direct flights: Frontier $149, United $189, Southwest $210. Which one?"
4
Budget Check Finance API
"Is $189 within my travel budget?"
check_budget(category="travel", amount=189) → $500 remaining
"You have $500 travel budget remaining. $189 is fine. Book the United flight?"
5
User Confirms Voice Input
"Yes, book it"
6
Book + Pay Travel API Payment
book_flight(flight_id="UA-1234")
charge_card(card_id="@vault:cw-pvt-ajay-default-card", amount=189)
Credit card token from vault — never exposed
7
Update Calendar Microsoft Graph
create_event(date="Saturday", title="Flight to Denver UA1234", time="7:00 AM")
8
Notify Teams / Slack
"All done! United flight to Denver Saturday 7AM, $189 charged, calendar updated. Confirmation sent to Teams."

What Makes This Possible in ContextWeaver

CapabilityHow It's Used
Multi-connector pluginOne plugin wires together Calendar + Flights + Finance + Payment + Notifications
Vault-backed secretsCredit card token, API keys stored in vault — user never sees them, plugin code never embeds them
Private scopeEach user's card, calendar, and budget are in their private connector extension — no one else can access them
Agentic loopThe AI doesn't just search — it reasons across steps, detects conflicts, asks for decisions, then executes
MCP ToolsEach action (check_calendar, search_flights, charge_card) is a registered MCP tool the AI can invoke
MCP Promptsbook-travel prompt template guides the multi-step flow
VoiceAny voice-to-text frontend (Alexa, Siri, web speech API) sends text to the MCP chat — ContextWeaver handles the rest

Can Anyone Else Do This Today?

Platform Verdict Why Not
ChatGPT + Plugins ⚠️ Partial No credential hierarchy, no private scoping, no vault. Each plugin is isolated — can't chain Calendar, Flights, Pay, Calendar in one flow
Claude + MCP ⚠️ Partial MCP tools yes, but single-user, no RBAC, no vault, no connector hierarchy — you'd hardcode API keys
Siri / Google Assistant ⚠️ Limited Pre-built actions only. Can't add custom connectors, no enterprise RBAC, no budget integration
Alexa Skills ⚠️ Limited Each skill is siloed. Can't chain across different services in one conversation
Microsoft Copilot ⚠️ Partial Can access M365 but can't easily wire in custom travel APIs + payment + budget checks
ContextWeaver ✅ Full Flow Multi-connector plugin, vault-backed secrets per user, agentic reasoning loop, enterprise RBAC, all in one conversation

The key differentiator: No other platform lets you visually compose a plugin that chains Calendar + Flights + Finance + Payment + Notifications with per-user vault-backed credentials and team-scoped access control — all deployable across any cloud with zero code.

Connectors Used in This Scenario

📅
Microsoft Graph
Calendar read/write
🔒 OAuth token in vault
✈️
Travel API
Amadeus / Sabre
🔒 API key in vault
💳
Finance API
Plaid / Expense system
🔒 Access token in vault
💳
Payment
Stripe / saved card
🔒 Card token in vault
💬
Teams / Slack
Notifications
🔒 Webhook in vault
All secrets stored in vault with per-user private scope — no credentials in code, config, or browser.
Back to About