Personal memory you can see and manage
You can now view and manage your memory with a full management interface on web and desktop, enabling you to edit your memory with the help of Computer.
π For more information about Computer by DevRev, refer to Computer by DevRev: My Computer: Personal Computer agent| DevRev
CX Agent Builder generates a working agent from past conversations
Building a production support agent used to take weeks of manual setup. The CX Agent Builder now ingests a customer's past conversations to auto-generate intents, guardrails, tone, and escalation rules, which a human reviews and tweaks before shipping, cutting cold-start time dramatically.
π For more information about Computer for your customers, refer to Computer for Your Customers: AI Self-Service Support | DevRev.
Voice agents connected to your telephony provider
You can now connect voice-based agents to your telephony provider so callers can have real-time conversations with an AI agent over the phone. Voice agents use the agent's existing context, skills, and tools during the call, bringing the same intelligent assistance to your phone channel.
More reliable voice support at scale
Voice support saw major concurrency and reliability work, including memory-aware admission control, more reliable call teardown, right-sized capacity, and expanded regional coverage. Voice calls are steadier under load and start with proper context, and traces are more accurate for debugging.
Fewer fallback and greeting glitches in agent responses
We fixed cases where the agent intermittently returned a default fallback message instead of a contextual answer, along with repeated affirmation and greeting behavior. Customer-facing responses are more consistent.
More accurate SLA and OLA behavior
Service and operational level agreements no longer trigger incorrectly after a customer comments on a closed ticket, and OLAs stop running on closed tickets. We also fixed a blank SLA policy page on reload and corrected an edit dialog that mislabeled OLA policies as SLA.
π For more information about service level agreements, refer to Service-level agreements.
More trustworthy CSAT surveys
CSAT now prevents multiple survey submissions per dispatch and adds a workflow-based trigger so surveys fire exactly when intended. Survey results are cleaner and easier to trust.
π For more information about CSAT, refer to CSAT on ticket.
Richer WhatsApp messaging for support
The WhatsApp channel moves toward parity with WhatsApp Business messaging. Outbound messages now support interactive quick-action reply and CTA buttons, with the ability to build workflows that branch on a customer's button response for dynamic multi-step flows. Templates and messages gain multi-variable support for dynamic placeholders like customer name and ticket ID, plus rendering of configured template headers and footers.
MCP connectors on mobile
AirSync now supports MCP connectors on Android and iOS. You can discover your org's provisioned servers in a searchable marketplace, add connectors during onboarding, and manage connections on the go, including resyncing, disconnecting, editing, and removing servers or agent connections.
Improved semantic search
Reranking is now enabled for all search queries, improving result relevance. Improved data extraction gives Computer the context it needs upfront, reducing the number of search calls needed to reach an answer faster.
Custom webhooks as a workflow trigger
Workflows can now be triggered by webhooks from external systems like Stripe, GitHub, and PagerDuty, without building an intermediary service. Webhooks are reusable named objects that many workflows can share, with a 30-second test capture, event history across workflows, jq and rule-builder filtering, and multiple authentication modes.
π For more information about webhook triggers, refer to Webhook Triggers.
Undo and redo, workflow tags, and child-part triggers
Workflow Builder gained undo and redo for step create, delete, and link deletion, workflow tags enabled for all orgs, and the ability to include child parts in workflow nodes. Building and organizing workflows is faster and less error-prone.
π For more information about workflows, refer to Workflows overview.
Connect any MCP server to your agents
DevRev now works as an MCP client, with a complete flow for adding MCP servers from the marketplace or via a custom URL and installing them to an agent. Agents can use a much wider set of external tools.
Stronger agent evaluation and bulk testing
Agent Evals got a round of fixes for bulk testing and observe dashboards, including reliable dataset re-imports, structured error reporting like rate limits, and correct titles in bulk test reports. Evaluating agents at scale is more dependable.
IP Address and Geolocation on Session Replays
Session replays can now be enriched with network-origin context. On session start, the ingestion service derives the client IP from the request and runs an offline reverse-IP lookup to attach country, region, city, and coordinates to the session record, with no SDK change required. The capability is fully opt-in and off by default, with independent controls for raw IP capture and geo enrichment that customers enable per org via a toggle in PLuG settings. Teams can now slice and filter sessions by location to understand where users are and speed up debugging.