SANCE is an API-first platform for AI agents that hold real conversations — both outbound sales and customer service, across all major messengers. The system is natively multi-agent, shares memory across agents, is grounded in your own knowledge base, and is driven by a configurable flow engine with its own dispatch logic, tools, and hooks. Over 20 distinct sub-agent types take part in every generation.
SANCE — the API for AI sales & service agents
AI agents that run your sales and customer-service conversations across every channel — and a funnel that improves itself.
And it's more than automation — it's the scientific method, built in. SANCE turns conversations into data, patterns into hypotheses, hypotheses into live A/B experiments, and winning variants into production — data → hypothesis → experiment → deployment — a closed loop that runs natively on the platform.
How SANCE works
The platform runs every conversation — inbound and outbound — through one pipeline, then learns from the outcomes to improve itself.
1
Messages & triggers
Inbound messages, files, and events — or an outbound campaign — kick off the pipeline.
WhatsAppLinkedIn10+ more
2
Orchestration
A visual state machine drives each turn: rules, classifiers, and LLM-judged decisions.
state treeclassifierssanitization
3
Context assembly
Shared memory, CRM, files, and your knowledge base (RAG) come together as one context.
shared memoryCRMRAG
4
Agentic pipeline
20+ specialized sub-agents produce a checked, reliable, and structured result.
quality checkroutingstate check
5
Execution
Reply, update CRM, book meetings, move stages, or hand off to a human or another agent.
replyactionsescalation
6
Learn & self-optimize
The system proposes data-driven A/B experiments that you approve in one click.
metricsA/BRL
Self-optimization loopdata→hypothesis→experiment→deploy
Platform capabilities
Agent Runtime
A visual state machine with 20+ sub-agents, hooks, tools, and webhooks — running inbound and outbound conversations.
Omnichannel Conversations
Two-way messaging on WhatsApp, Telegram, LinkedIn, email, SMS, and more — voice, images, and files included.
Shared Memory
Conversations, actions, tool calls, and the business outcomes of every interaction with the lead are available to every agent.
Outbound & Dispatch
Multi-step campaigns with lead research, follow-up pings, contact rotation, and deliverability controls.
Knowledge Base
In-house RAG — answers grounded in your documents, files, and CRM data.
Experiment Loop
The scientific method, built in: the platform proposes the next experiment and rewrites the prompt — you approve in one click.
CRM Sync
Two-way sync with amoCRM, Kommo, Bitrix24, and more — leads, stages, fields, tags, and messages.
Human clarification
Human-in-the-loop: when confidence is low the agent asks an operator, continues with their answer, and saves it to its knowledge base so it never has to ask again.
Files & Media
Send and receive voice (auto-transcribed), images, and documents in any conversation.
Meetings & Calendar
Agents book meetings directly through your calendar integrations.
Funnel Analytics
Funnel analytics with per-dialogue LLM cost attribution.
Scheduling & Timing
Working-hours scheduling, human-like typing delays, message splitting, and debounce.
Redirect
A lead asks you to message a different number — and the agent does it, with the full context.
Input Sanitization
A two-pass check screens every inbound message for prompt injection and abuse before it reaches the model.
Dialogue Resumer
No conversation is left unanswered — fallback chains across multiple LLM providers retry until a reply goes out.
PII Redaction
Strip personal data from requests before they are sent to LLM providers.