> ## Documentation Index > Fetch the complete documentation index at: https://docs.bland.ai/llms.txt > Use this file to discover all available pages before exploring further. # July 21, 2026 > Agent speech timing controls, Scheduling node status routing, and full REST support for Custom API tools ### Agent Speech Timing: Interruptibility & Resumption Speed Two controls now shape your agent's conversational timing: how readily it yields when a caller talks over it, and how quickly it responds once the caller stops. Set them per node, on inbound numbers, or in the [Send Call API](https://docs.bland.ai/api-v1/post/calls); see the [Agent Speech docs](/tutorials/agent-speech) for a full guide. * Interruptibility now spans five levels, from "Block interruptions" up to a new "Always interrupt", with the middle levels relabeled Easy, Balanced, and Difficult * The new Resumption Speed setting (Slow, Medium, Fast) sets how long the agent waits after the caller finishes before responding * Node-level settings persist for the rest of the call unless overridden; defaults are unchanged Set agent interruptibility node setting with five levels from Block interruptions to Always interrupt *** ### Status Code Routing on the Scheduling Node Scheduling nodes now support the same deterministic "Pathway After API Request Response" routing as Webhook nodes. * Route to a specific node based on the response status code or extracted response fields * Unmatched responses keep the caller in the node, so the booking flow continues *** ### Custom API Tools: PUT, PATCH, and DELETE The Custom API tool builder now supports all five REST verbs, so tools can update and delete records in external systems, not just read and create them. * PUT, PATCH, and DELETE work across tool creation, editing, test runs, and cURL import * JSON body validation errors surface before you save instead of failing on a live call Custom API tool request method dropdown showing GET, POST, PUT, PATCH, and DELETE *** ### Improvements **Pathways** * Pathway logs label node tool calls as "Tool Call" with the tool's name instead of a generic "Webhook" entry, and unmet loop conditions show "Condition not met" instead of a red failure * Node tools trigger reliably when the agent drops a tool's trailing annotation tag, and unmatched tool calls are logged as clear errors instead of failing silently * The Reset STT Language node is now available to all organizations **Evals** * Large [eval](https://app.bland.ai/dashboard/evals) runs now complete reliably at any size, and verdicts display for every call in runs larger than 100 calls **Telephony** * \[Enterprise] Talkdesk and NICE CXone are now supported BYOC SIP providers **Organization & Security** * Invite teammates by email as well as phone; invitees get a sign-up link and join your organization automatically * \[Enterprise] SSO providers can now be edited in place, so you can rotate an expiring client secret or certificate without recreating the provider. **Billing** * Checkout on the [billing page](https://app.bland.ai/dashboard/pay) now runs inline with clearer purchase errors, you can manage or cancel individual phone number subscriptions directly, and double-clicks can no longer trigger duplicate purchases **Personas** * Personas can now define voicemail behavior: hang up, ignore, leave a message, or send an SMS **Automations** * The Send Call action within automations can attach custom metadata to calls, included in the post-call webhook for every status (including no-answer) so you can correlate webhooks to your records **Messaging** * SMS log exports now include a metadata column **Agent Testing** * Run a scenario N times to see a success rate, and runs that fail before the first turn get a "Failed to start" badge