Changelog
v5.22.3 — 2026-06-01 🚀 SCIM, Webhooks & OpenAPI
Features
- SCIM 2.0 Provisioning API (
/v1/scim/*): Added full support for SCIM 2.0 provisioning to manage Users and Groups via Identity Providers (Azure AD, Okta, Ping). Includes endpoints forUsersandGroupswithGET,POST,PUT,PATCH, andDELETE. - Outbound Webhooks: Added real-time event subscriptions via webhooks with HMAC-SHA256 signature verification (
X-Senda-Signature). Events includeagent.action.executed,chat.session.started, andknowledge.document.ingested. - OpenAPI Interactive Explorer: Released the new Swagger UI at
/api/docs/uiand raw OpenAPI JSON spec at/api/docs, integrated directly into the portal for live testing.
Improvements
- Chat History API: Added
historyparameter toPOST /v1/chatto allow developers to supply past conversational context directly in stateless API interactions.
v3.0.0 — 2026-05-07 🚀 Senda Platform API
Phase A: Foundation
WS-1: Public REST API
- API Key authentication (SHA-256 hashed, scope-based RBAC)
- 7 public endpoints (
/v1/chat,/v1/agents,/v1/knowledge,/v1/actions,/v1/usage,/v1/keys) - KV-based rate limiting with
X-RateLimitheaders
WS-2: Embeddable Widget
<senda-chat>Web Component with Shadow DOM (12.2 kB, 4 kB gzip)- CSS custom property theming system
senda:messageevent for host page integration
WS-3: MCP Server
- Full JSON-RPC 2.0 implementation (MCP spec 2025-03-26)
- 5 tools:
senda_chat,senda_search_knowledge,senda_execute_action,senda_list_agents,senda_ingest_document - 3 resources:
senda://agents,senda://knowledge/{id},senda://actions
Phase B: Extension
WS-4: MCP Client Proxy
- Action Engine now supports
action_type = 'mcp'for consuming third-party MCP servers - Full audit logging for MCP tool calls
WS-5: Chrome Extension
- Manifest V3 with side panel and content script
- Contextual awareness (page URL, title, selected text)
WS-7: npm SDK
@senda/sdk— Fully typed TypeScript SDK- Async iterator streaming (
for await...of) - Zero dependencies
v2.6.1 — 2026-05-07
Fixed
- Router Anti-Confusion: Negative examples prevent incorrect agent assignment
- Hallucination Post-Filter: Detects false "Action Executed" claims during slot filling
- Conflict Resolution: Consolidation prompt resolves contradictory learnings