Raw, Capped AI Agent Pricing: Why Embage Avoids Per-Session Flat Fees
How Embage keeps AI agent costs predictable by separating model budgets from capped monthly action quotas (knowledge base searches, CRM writes, and integrations).
Most AI chatbot vendors make pricing appear deceptively simple by charging flat per-session or per-conversation fees. At first glance, a "$0.50 per chat" price tag sounds straightforward. In practice, per-session pricing is a trap that penalizes businesses for basic customer inquiries while creating unpredictable monthly bills.
Embage takes a fundamentally different approach. We do not bill chat as a single flat session. Instead, Embage plans combine a model usage budget (for credit-based LLM selection) with capped monthly action quotas for the specific tasks an agent actually completes.
The Flaw in Flat Per-Session Chat Pricing
Per-session pricing obscures the true infrastructure cost of an AI interaction. Consider two completely different website visitors:
- Visitor A (Simple FAQ): Asks "What are your store hours?" The agent responds in 8 words using 50 tokens.
- Visitor B (Complex Support Investigation): Engages in a 20-message troubleshooting session requiring three vector knowledge base searches, querying an external order database, creating a CRM support ticket, and dispatching a Gmail follow-up.
Under flat per-session pricing, your business pays the exact same high rate for both interactions. The tiny FAQ question subsidizes the complex enterprise workflow. Worse still, if a bot or viral surge triggers thousands of simple 1-message chats, your monthly bill explodes unexpectedly.
How Embage Pricing Works: Raw Usage + Capped Actions
Embage decouples conversation handling into transparent, measurable operational components:
+-----------------------------------------------------------------------+
| Embage Monthly Operating Envelope |
+-----------------------------------------------------------------------+
| |
+--> 1. Model Provider Budget (Token usage for Gemini 2.0 Flash / LLMs)
|
+--> 2. Knowledge Base Search Cap (Vector retrieval operations)
|
+--> 3. Datastore Write Cap (CRM lead, ticket, feedback, & review logs)
|
+--> 4. Integration Call Cap (3rd-party SaaS execution)
|
+--> 5. Voice Minute Allotment (Real-time sub-second voice WebSockets)
1. Model Provider Budget
Covering raw LLM token consumption. Because Embage lets you choose performance-optimized models (like Gemini 2.0 Flash), standard chat interactions consume minimal token budget.
2. Knowledge Base Search Quota (KNOWLEDGE_BASE_SEARCH)
Actions are billed only when an agent invokes the knowledge_base sub-agent tool to query your uploaded PDFs, text documentation, or web vectors. Standard conversational greetings do not touch this cap.
3. Datastore Write Quota (DATASTORE_WRITES)
Actions count only when an agent executes the add_to_datastore tool to write structured records into your organization's built-in CRM Datastores (such as support tickets, qualified sales leads, product reviews, or customer feedback).
4. Integration Call Quota (INTEGRATION_CALLS)
Actions count when a sub-agent executes the integration_3p tool to interact with connected third-party SaaS platforms (e.g. Gmail, Slack, GitHub, Google Calendar, HubSpot, Salesforce, or Shopify).
5. Voice-to-Voice Minute Allotments
Real-time voice interactions utilize sub-second WebSockets audio streaming. Embage plans include dedicated monthly voice-minute allotments. Tool calls made during a voice call (such as searching knowledge or writing a CRM record) consume from the same transparent action caps.
A Side-by-Side Comparison
Let's compare how two interactions are billed in Embage:
| Interaction Type | Actions Triggered | Embage Billing Impact |
|---|---|---|
| Simple FAQ Chat "Are you open on weekends?" |
• None | • Consumes standard model token budget. • 0 Action Caps consumed. |
| Full Support Workflow "I was double billed on invoice #8812" |
• 1 Knowledge Search • 1 Datastore Ticket Write • 1 Gmail Integration Call |
• Consumes model token budget. • 1 Knowledge Search Cap. • 1 Datastore Write Cap. • 1 Integration Call Cap. |
Why Hard Monthly Caps Protect Your Business
In Embage, limits are hard monthly caps, not hidden overage penalties.
If your organization reaches its monthly quota for a specific tool (for example, INTEGRATION_CALLS), the specific integration tool pauses safely until the next billing cycle or explicit tier upgrade. The agent continues answering questions using the Knowledge Base and remaining available tools without interruption.
This design guarantees two major operational benefits:
- Predictable Expenses: No surprise $5,000 credit card bills from unexpected traffic spikes.
- Value-Aligned Billing: You only pay for high-value business outcomes (leads captured, tickets logged, emails sent) rather than idle chit-chat.
Further Reading
- Explore Embage Platform Features — Knowledge base, CRM datastores, sub-agents, and integrations.
- AI Agent Orchestration for Customer Support — How sub-agents minimize token usage.
- Gmail AI Agent Workflows — Automated follow-up setup guide.
- Shopify AI Agent Product Lookup — E-commerce support automation.