Agentic Index: agent infrastructure platforms change log https://agenticindex.io/rankings/agent-infrastructure Verified changes to agent infrastructure platforms, from the Agentic Index change log. Every entry is dated and sourced from public vendor material. en-us Free to reuse with a visible link to agenticindex.io. Terms: https://agenticindex.io/use-this-data Fri, 25 Sep 2026 12:00:00 GMT 720 https://media.base44.com/images/public/69e4cec10b4126d233ee7ee4/a44c0d34a_generated_image.png Agentic Index: agent infrastructure platforms change log https://agenticindex.io/rankings/agent-infrastructure LiteLLM launched the public preview of LiteAgents, an SDK that supports Deep Agents, Pydantic AI, Claude Agent SDK, Codex and OpenCode… https://agenticindex.io/change-log/litellm agenticindex.io:change:litellm-k9gz3u-2026-09-25 Fri, 25 Sep 2026 12:00:00 GMT MCP / tool calling / API Agent infrastructure LiteLLM launched the public preview of LiteAgents, an SDK that supports Deep Agents, Pydantic AI, Claude Agent SDK, Codex and OpenCode through a common query interface. Profiles select the runtime while retaining shared tools, MCP connections and model configuration; native runtime options remain available. Optional Temporal integration adds recovery from recorded operations and checkpoints, while simple local agents require neither Temporal nor PostgreSQL. Changing runtimes starts new runs and does not migrate an active native session.

Why it matters: Teams can compare agent runtimes on their own workloads with less integration work and add recovery for longer jobs. Production evaluation should account for preview maturity and ensure that tools causing external changes tolerate retries.

Impact: High. Verification: Verified. Type: MCP / tool calling / API.

Evidence: Official blog

Source: Agentic Index change log. Free to reuse with a link back. Terms.

]]> Windmill v1.818.0 counts enabled service accounts as operator seats in self hosted Enterprise deployments, billing each as half a seat… https://agenticindex.io/change-log/windmill agenticindex.io:change:windmill-170vtl-2026-09-25 Fri, 25 Sep 2026 12:00:00 GMT Pricing / packaging Agent infrastructure Agent builder Windmill v1.818.0 counts enabled service accounts as operator seats in self hosted Enterprise deployments, billing each as half a seat regardless of its assigned role. Enabled service accounts also count toward the offline license seat cap, and creation is refused when the cap is reached. The related Cloud Team billing change is separate from this release.

Why it matters: Organizations using dedicated identities for agents and automation need to include those accounts in license forecasts. Reviewing unused enabled accounts before upgrading can help avoid unexpected charges or blocked account creation.

Impact: High. Verification: Verified. Type: Pricing / packaging.

Evidence: GitHub release

Source: Agentic Index change log. Free to reuse with a link back. Terms.

]]> Exa launched Agent Ultra, its highest effort mode for exhaustive research and building large lists of entities. https://agenticindex.io/change-log/exa agenticindex.io:change:exa-zqk2ll-2026-09-25 Fri, 25 Sep 2026 12:00:00 GMT Agent capability Agent infrastructure Exa launched Agent Ultra, its highest effort mode for exhaustive research and building large lists of entities. It coordinates agents across many sources and supports research runs lasting up to three hours. Users can set spending and duration limits or stop a run early while retaining the results gathered so far.

Why it matters: Research teams can allocate more computation to difficult verification and enrichment tasks while controlling the budget for each run. This offers a configurable option for work where completeness matters more than response speed.

Impact: Medium. Verification: Verified. Type: Agent capability.

Evidence: Official blog

Source: Agentic Index change log. Free to reuse with a link back. Terms.

]]> Browserbase Agents can pause a run for human input and resume within the same browser session. https://agenticindex.io/change-log/browserbase agenticindex.io:change:browserbase-1jwxbb-2026-09-24 Thu, 24 Sep 2026 12:00:00 GMT Human approval / guardrails Browser / computer-use agent Agent infrastructure Browserbase Agents can pause a run for human input and resume within the same browser session. Developers set a natural language pauseWhen condition, which gives the agent a pause tool and lets it enter PAUSED when the condition is met. The resume API accepts a follow up task, such as continuing after a user supplies an MFA code.

Why it matters: Teams can include human checkpoints in browser workflows without rebuilding the session after each interruption. They should test that their pause conditions reliably identify the situations requiring human input.

Impact: High. Verification: Verified. Type: Human approval / guardrails.

Evidence: Official changelog

Source: Agentic Index change log. Free to reuse with a link back. Terms.

]]> Corti updated Guided Generation to pass user edited, added, and discarded facts with explicit type metadata. https://agenticindex.io/change-log/corti agenticindex.io:change:corti-qgp3j2-2026-09-24 Thu, 24 Sep 2026 12:00:00 GMT Human approval / guardrails Healthcare agent Agent infrastructure Corti updated Guided Generation to pass user edited, added, and discarded facts with explicit type metadata. User edits now take precedence over conflicting transcript content, and discarded facts are excluded from generated notes. User added facts and system supplied context also receive explicit priority.

Why it matters: Clinical application teams should test correction and deletion workflows against this release, especially where generated notes previously reintroduced rejected transcript details.

Impact: Medium. Verification: Verified. Type: Human approval / guardrails.

Evidence: Official changelog

Source: Agentic Index change log. Free to reuse with a link back. Terms.

]]> DBOS Python 3.1.0 added workflow rewind, allowing a workflow in a terminal state to restart from a specified step. https://agenticindex.io/change-log/dbos agenticindex.io:change:dbos-gfcvc0-2026-09-24 Thu, 24 Sep 2026 12:00:00 GMT Workflow orchestration Agent infrastructure DBOS Python 3.1.0 added workflow rewind, allowing a workflow in a terminal state to restart from a specified step. The release also added workflow ID reuse policies and local configuration of Conductor metadata only mode.

Why it matters: Teams operating durable agent workflows gain a recovery option for completed or failed runs. Rewind can repeat downstream messages and actions, so recovery procedures should account for those effects before replaying a run.

Impact: Medium. Verification: Verified. Type: Workflow orchestration.

Evidence: GitHub release

Source: Agentic Index change log. Free to reuse with a link back. Terms.

]]> LangSmith launched the public beta of smithtune, a CLI and coding agent skill that turns recorded agent trajectories into supervised fine… https://agenticindex.io/change-log/langsmith agenticindex.io:change:langsmith-1m7p56-2026-09-24 Thu, 24 Sep 2026 12:00:00 GMT Workflow orchestration Agent infrastructure LangSmith launched the public beta of smithtune, a CLI and coding agent skill that turns recorded agent trajectories into supervised fine tuning workflows. It prepares and persists training datasets, submits training jobs to Fireworks or Baseten, compares the resulting model with the base model through LangSmith evaluations, and can deploy the selected model. Users need a LangSmith account with agent traces and an API key for a supported training provider.

Why it matters: Teams can test whether their production examples support a specialized agent model without building separate data preparation, training, and evaluation pipelines. Recorded datasets and evaluation comparisons give reviewers a way to inspect gains and regressions before deployment.

Impact: High. Verification: Verified. Type: Workflow orchestration.

Evidence: Official blog

Source: Agentic Index change log. Free to reuse with a link back. Terms.

]]> Braintrust announced Nitro, a new asynchronous query engine for Brainstore that separates storage reads from compute work to accelerate… https://agenticindex.io/change-log/braintrust agenticindex.io:change:braintrust-ok05u5-2026-09-24 Thu, 24 Sep 2026 12:00:00 GMT Observability / auditability Agent infrastructure Braintrust announced Nitro, a new asynchronous query engine for Brainstore that separates storage reads from compute work to accelerate trace investigations. Nitro is automatically enabled for SaaS and BYOC customers, with support starting at dataplane 2.15 for deployments customers host themselves. Braintrust reports that full text searches ran more than twice as fast in its production comparison.

Why it matters: Teams investigating agent failures can search large trace histories with less waiting, including when coding agents issue many concurrent queries. Existing hosted customers receive the improvement without configuring a new search interface.

Impact: Medium. Verification: Verified. Type: Observability / auditability.

Evidence: Official blog

Source: Agentic Index change log. Free to reuse with a link back. Terms.

]]> Cognee v1.6.1 adds a Google Drive OAuth connector, Gmail ingestion, and shared drive synchronization, with Google connectors bundled in… https://agenticindex.io/change-log/cognee agenticindex.io:change:cognee-85mffc-2026-09-24 Thu, 24 Sep 2026 12:00:00 GMT Integrations Agent infrastructure Cognee v1.6.1 adds a Google Drive OAuth connector, Gmail ingestion, and shared drive synchronization, with Google connectors bundled in the SDK. The release also streams graph visualization responses and preserves document metadata in retrieved chunks while restricting provenance to datasets the caller can read.

Why it matters: Teams can populate agent memory from shared Google workspace content with fewer integration steps. Large memory graphs become easier to inspect, and retrieval retains clearer links to source documents.

Impact: Medium. Verification: Verified. Type: Integrations.

Evidence: GitHub release

Source: Agentic Index change log. Free to reuse with a link back. Terms.

]]> Agno 3.0.11 introduced a shared Knowledge retrieval pipeline with MMR reranking for result diversity and recency reranking to favor newer… https://agenticindex.io/change-log/agno agenticindex.io:change:agno-uyyzyp-2026-09-23 Wed, 23 Sep 2026 12:00:00 GMT MCP / tool calling / API Agent infrastructure Agno 3.0.11 introduced a shared Knowledge retrieval pipeline with MMR reranking for result diversity and recency reranking to favor newer revisions. It deprecates reranker configuration on individual vector databases in favor of the Knowledge layer. The release adds guarded APIs for relocating indexed documentation sources, with migration running as a dry run by default. Typed page tool results expose errors, completeness and truncation, while run outputs and API schemas gain a cancellation_stage field.

Why it matters: Teams can apply retrieval policies consistently across storage backends and distinguish incomplete evidence from complete results before an agent acts. Existing integrations should migrate reranker configuration and can replace brittle cancellation message parsing with explicit status data.

Impact: Medium. Verification: Verified. Type: MCP / tool calling / API.

Evidence: GitHub release

Source: Agentic Index change log. Free to reuse with a link back. Terms.

]]> Cartesia introduced Multilingual Voices, a library of more than 50 voices that support up to 25 languages while retaining the same vocal… https://agenticindex.io/change-log/cartesia agenticindex.io:change:cartesia-pcsges-2026-09-23 Wed, 23 Sep 2026 12:00:00 GMT Agent capability Agent infrastructure Voice agent Cartesia introduced Multilingual Voices, a library of more than 50 voices that support up to 25 languages while retaining the same vocal identity. Customers can keep a voice ID across its supported languages and select a locale in speech generation requests. The offering is available through the voice library and API, with language and accent support varying by voice.

Why it matters: Voice agent teams can expand language coverage without assigning a different speaker identity to each market. Buyers should validate each voice's supported locales and fallback behavior before deploying it across languages.

Impact: Medium. Verification: Verified. Type: Agent capability.

Evidence: Official blog

Source: Agentic Index change log. Free to reuse with a link back. Terms.

]]> Forest's September release lets Claude, Dust and other MCP clients start Forest Workflows through its MCP server. https://agenticindex.io/change-log/forest-admin agenticindex.io:change:forest-admin-17p4b5-2026-09-23 Wed, 23 Sep 2026 12:00:00 GMT Workflow orchestration Enterprise operations agent Agent infrastructure Forest's September release lets Claude, Dust and other MCP clients start Forest Workflows through its MCP server. Decision steps now offer Automatic (Deterministic) execution, evaluating conditions against preceding Get Data steps without model interpretation during execution. Manual, AI assisted and Full AI modes remain available for decisions that need interpretation.

Why it matters: Operations teams can make governed business processes available to external agents while keeping predictable routing under explicit rules. Buyers can distinguish where a model's judgment is necessary from the parts of a process that should behave consistently on every run.

Impact: Medium. Verification: Verified. Type: Workflow orchestration.

Evidence: Official changelog

Source: Agentic Index change log. Free to reuse with a link back. Terms.

]]> Arcade announced a new Snowflake toolkit that lets agents discover warehouses, databases, schemas, tables, and columns, then run queries… https://agenticindex.io/change-log/arcade agenticindex.io:change:arcade-rvnbry-2026-09-23 Wed, 23 Sep 2026 12:00:00 GMT Integrations Agent infrastructure Arcade announced a new Snowflake toolkit that lets agents discover warehouses, databases, schemas, tables, and columns, then run queries using the requesting user's identity. Existing Snowflake grants determine what the agent can access. The toolkit permits read only queries and blocks writes, schema changes, and role switches.

Why it matters: Enterprise teams can give agents warehouse access while retaining existing user permissions and data policies. The read only boundary limits mutation risk, though query compute costs still need operational controls.

Impact: Medium. Verification: Verified. Type: Integrations.

Evidence: Official blog

Source: Agentic Index change log. Free to reuse with a link back. Terms.

]]> Prime Intellect made Prime Sandboxes generally available through its CLI and SDK and as part of its reinforcement learning stack. https://agenticindex.io/change-log/prime-intellect agenticindex.io:change:prime-intellect-1fz57x-2026-09-23 Wed, 23 Sep 2026 12:00:00 GMT Deployment / data residency Agent infrastructure Prime Intellect made Prime Sandboxes generally available through its CLI and SDK and as part of its reinforcement learning stack. Each sandbox runs a Linux microVM with its own kernel, supporting Docker Compose, background services and custom Docker images. Accounts start with capacity for 1,024 concurrent sandboxes.

Why it matters: Teams training and evaluating agents can use VM isolation while retaining their existing Docker environment definitions. Native Docker and kernel support broaden the workloads they can test without operating the underlying VM fleet.

Impact: High. Verification: Verified. Type: Deployment / data residency.

Evidence: Official blog

Source: Agentic Index change log. Free to reuse with a link back. Terms.

]]> W&B Weave: W&B Server v0.85.0 adds links in both directions between Weave calls and the agent spans they produce or invoke. https://agenticindex.io/change-log/wandb-weave agenticindex.io:change:wandb-weave-lffuaj-2026-09-23 Wed, 23 Sep 2026 12:00:00 GMT Observability / auditability Agent infrastructure W&B Server v0.85.0 adds links in both directions between Weave calls and the agent spans they produce or invoke. The same release adds a configurable process wide query parallelism budget, allowing deployments to reduce peak memory use when Weave workloads load large text heavy tables by trading off request concurrency.

Why it matters: Agent operators can follow execution across call and span views when investigating behavior. Deployment owners can tune resource use for large analytical workloads instead of relying on the fixed default concurrency.

Impact: Medium. Verification: Partially Verified. Type: Observability / auditability.

Evidence: Official changelog

Source: Agentic Index change log. Free to reuse with a link back. Terms.

]]> FlowX.AI documented a Browser Automation node for Agent Builder, available from release 5.13.0 on SaaS. https://agenticindex.io/change-log/flowx-ai agenticindex.io:change:flowx-ai-b7s3kq-2026-09-23 Wed, 23 Sep 2026 12:00:00 GMT Agent capability Agent builder Enterprise operations agent Agent infrastructure Multi-agent platform FlowX.AI documented a Browser Automation node for Agent Builder, available from release 5.13.0 on SaaS. The node drives a real browser through web applications that have no API, filling multi page forms, navigating portals and retrieving results from a task written in plain language. It checks outcomes against the target site's recorded state rather than the agent's own report and keeps the browser session across workflow steps. Self hosted deployments receive it with the next LTS release family.

Why it matters: Banks and insurers running agents on FlowX can now reach legacy portals and third party sites that expose no integration point from inside the same governed workflow, with every run logged. Verification against the site's recorded state is the detail to test in a pilot, since an agent reporting success on a form it never submitted is the usual failure in this category.

Impact: High. Verification: Verified. Type: Agent capability.

Evidence: Official documentation

Source: Agentic Index change log. Free to reuse with a link back. Terms.

]]> Mastra released @mastra/mcp 2.0.0, whose servers support only the July 28, 2026 MCP revision and require @mastra/core 1.68 or newer. https://agenticindex.io/change-log/mastra agenticindex.io:change:mastra-4khz10-2026-09-22 Tue, 22 Sep 2026 12:00:00 GMT MCP / tool calling / API Agent infrastructure Mastra released @mastra/mcp 2.0.0, whose servers support only the July 28, 2026 MCP revision and require @mastra/core 1.68 or newer. Tools that need caller input now suspend and resume through input_required continuations with signed request state. The release removes legacy initialization, session headers, standalone HTTP plus SSE transport, and earlier elicitation APIs.

Why it matters: Teams upgrading existing MCP integrations need to revise transport and input handling code and verify client compatibility. Deployments serving older clients can retain supported 1.x servers alongside 2.x servers.

Impact: High. Verification: Verified. Type: MCP / tool calling / API.

Evidence: GitHub release

Source: Agentic Index change log. Free to reuse with a link back. Terms.

]]> OpenRouter launched a Batch API for asynchronous chat completions, Responses, Messages, and embedding workloads across more than 70 models. https://agenticindex.io/change-log/openrouter agenticindex.io:change:openrouter-167fg4-2026-09-22 Tue, 22 Sep 2026 12:00:00 GMT MCP / tool calling / API Agent infrastructure OpenRouter launched a Batch API for asynchronous chat completions, Responses, Messages, and embedding workloads across more than 70 models. Developers submit an inline request array and poll for results within a 24 hour completion window, with token pricing typically discounted by 50%. Each batch runs on one eligible provider, and inputs and results remain stored for 30 days unless deleted.

Why it matters: Evaluation, labeling, and backlog processing jobs can trade response latency for lower inference cost through one interface. Teams should check provider eligibility and retention requirements before moving sensitive or urgent workloads into batch processing.

Impact: High. Verification: Verified. Type: MCP / tool calling / API.

Evidence: Official blog

Source: Agentic Index change log. Free to reuse with a link back. Terms.

]]> Smallest AI added POST /waves/v1/auth/token so servers can issue temporary credentials for browser and mobile speech clients while… https://agenticindex.io/change-log/smallest-ai agenticindex.io:change:smallest-ai-1gpsnt-2026-09-22 Tue, 22 Sep 2026 12:00:00 GMT MCP / tool calling / API Voice agent Agent infrastructure Customer support agent Smallest AI added POST /waves/v1/auth/token so servers can issue temporary credentials for browser and mobile speech clients while retaining the API key. Tokens last 30 to 900 seconds, defaulting to 300, and authorize supported speech synthesis, transcription, speech to speech, and voice listing routes. They cannot authorize voice cloning, LLM chat completions, analytics, or further token creation; usage remains billed to the issuing key.

Why it matters: Teams can expose speech inference to clients with narrower credentials instead of distributing a persistent API key. Integrators must use the same region for token issuance and inference, and account for open WebSocket sessions continuing after token expiry.

Impact: Medium. Verification: Verified. Type: MCP / tool calling / API.

Evidence: Official changelog

Source: Agentic Index change log. Free to reuse with a link back. Terms.

]]> TrueFoundry 0.174.0 enables plan based feature gating for nonpaying tenants on managed control planes. https://agenticindex.io/change-log/truefoundry agenticindex.io:change:truefoundry-sxsc3s-2026-09-22 Tue, 22 Sep 2026 12:00:00 GMT Pricing / packaging Agent infrastructure TrueFoundry 0.174.0 enables plan based feature gating for nonpaying tenants on managed control planes. Resource caps block further creation, and capabilities including agents, MCP tool approval, custom roles and SSO require higher plans. Existing configurations and traffic continue, but gated features disappear from the Developer plan UI; paying customers and self hosted or dedicated control planes are unaffected.

Why it matters: Teams using the Developer plan for a pilot must account for paid access when evaluating governance controls and expanding agent workloads. A successful existing deployment does not guarantee that its configuration remains manageable under the same plan.

Impact: High. Verification: Verified. Type: Pricing / packaging.

Evidence: Official documentation

Source: Agentic Index change log. Free to reuse with a link back. Terms.

]]> Firecrawl launched Alexandria, a common interface for agents to discover and retrieve information from official data providers, custom… https://agenticindex.io/change-log/firecrawl agenticindex.io:change:firecrawl-3n8fuc-2026-09-22 Tue, 22 Sep 2026 12:00:00 GMT MCP / tool calling / API Agent infrastructure Firecrawl launched Alexandria, a common interface for agents to discover and retrieve information from official data providers, custom connectors and its Research, Developer and Government indexes. Agents can query providers and use specialized tools alongside live web search through Firecrawl.

Why it matters: Teams can expand an agent's research coverage without maintaining a separate connection and data format for each provider.

Impact: High. Verification: Verified. Type: MCP / tool calling / API.

Evidence: Official blog

Source: Agentic Index change log. Free to reuse with a link back. Terms.

]]> LangChain: LangGraph 1.2.12 added the optional response_schema parameter to interrupt(). https://agenticindex.io/change-log/langchain agenticindex.io:change:langchain-1c8vmm-2026-09-21 Mon, 21 Sep 2026 12:00:00 GMT Human approval / guardrails Agent infrastructure Multi-agent platform LangGraph 1.2.12 added the optional response_schema parameter to interrupt(). It exposes the expected response as JSON Schema and validates resumed values when supplied with a Pydantic model, TypedDict, or dataclass. Raw JSON Schema dictionaries are passed through without validation.

Why it matters: Teams building human approval workflows gain a more explicit contract at decision points, helping keep malformed responses out of persisted agent state.

Impact: Medium. Verification: Verified. Type: Human approval / guardrails.

Evidence: GitHub release

Source: Agentic Index change log. Free to reuse with a link back. Terms.

]]> Sourcegraph Cloud updated Code Finder to accept the repository as a separate parameter and expanded read_live_changesets so Agentic Batch… https://agenticindex.io/change-log/sourcegraph agenticindex.io:change:sourcegraph-2k6okp-2026-09-21 Mon, 21 Sep 2026 12:00:00 GMT MCP / tool calling / API Agent infrastructure Coding agent Sourcegraph Cloud updated Code Finder to accept the repository as a separate parameter and expanded read_live_changesets so Agentic Batch Changes can inspect live CI checks and branch state before requesting help. The same release adds structured publish credential problems to prepare_batch_spec metadata and enables repository fetch history logs. Deep Search gains safe date functions for analyzing commit activity and time windows.

Why it matters: Teams integrating code agents should review tool argument handling and can give agents more evidence before escalating failed changes to an engineer. Repository fetch history also helps administrators diagnose whether stale source data affected an investigation.

Impact: Medium. Verification: Verified. Type: MCP / tool calling / API.

Evidence: Official changelog

Source: Agentic Index change log. Free to reuse with a link back. Terms.

]]> Vapi added a serviceTier setting that lets assistants request faster OpenAI processing for supported GPT-5.6 and GPT-5.5 models. https://agenticindex.io/change-log/vapi agenticindex.io:change:vapi-14pxd8-2026-09-21 Mon, 21 Sep 2026 12:00:00 GMT MCP / tool calling / API Agent infrastructure Customer support agent Voice agent Vapi added a serviceTier setting that lets assistants request faster OpenAI processing for supported GPT-5.6 and GPT-5.5 models. Its September 21 update also adds provider request IDs to call logs for language model, transcription, and voice services when customers bring their own keys.

Why it matters: Voice teams can test a latency and cost tradeoff without changing their agent workflow, while request IDs make provider support investigations easier to correlate with specific calls. Fast processing uses the provider's corresponding rates and may be downgraded under load.

Impact: Medium. Verification: Verified. Type: MCP / tool calling / API.

Evidence: Official changelog

Source: Agentic Index change log. Free to reuse with a link back. Terms.

]]> Temporal introduced Amazon Bedrock AgentCore Runtime as a compute provider for Serverless Workers in prerelease. https://agenticindex.io/change-log/temporal agenticindex.io:change:temporal-zc0scx-2026-09-21 Mon, 21 Sep 2026 12:00:00 GMT Integrations Agent infrastructure Temporal introduced Amazon Bedrock AgentCore Runtime as a compute provider for Serverless Workers in prerelease. Teams can try the integration through self hosted Temporal, with workers started as task demand rises and workflow progress retained across worker failures. Temporal Cloud access remains subject to approval.

Why it matters: Teams building agents on AWS can evaluate elastic worker capacity without maintaining a continuously running fleet. Buyers should distinguish the publicly usable self hosted path from the restricted Cloud prerelease when planning adoption.

Impact: Medium. Verification: Verified. Type: Integrations.

Evidence: Official blog

Source: Agentic Index change log. Free to reuse with a link back. Terms.

]]> Confident AI launched Voice AI Evals with caller simulations covering interruptions, background noise, and turn patterns. https://agenticindex.io/change-log/confident-ai agenticindex.io:change:confident-ai-8swabv-2026-09-21 Mon, 21 Sep 2026 12:00:00 GMT Observability / auditability Agent infrastructure Confident AI launched Voice AI Evals with caller simulations covering interruptions, background noise, and turn patterns. Seven deterministic audio metrics assess naturalness, intelligibility, consistency, turn taking, responsiveness, audio integrity, and reliability alongside existing conversation evaluations. Supported connections include SIP, phone calls, WebRTC, WebSockets, and LiveKit integrations.

Why it matters: Teams evaluating voice agents can test audible behavior and conversational correctness in a shared workflow. This exposes failures that transcript scoring alone misses, such as poor interruption handling or unusable audio.

Impact: High. Verification: Verified. Type: Observability / auditability.

Evidence: Official blog

Source: Agentic Index change log. Free to reuse with a link back. Terms.

]]> Delight.ai introduces Intelligent Voice Handoff, which keeps a caller connected while the agent reaches and briefs the person taking over. https://agenticindex.io/change-log/sendbird agenticindex.io:change:sendbird-x0rpe6-2026-09-21 Mon, 21 Sep 2026 12:00:00 GMT Workflow orchestration Customer support agent Agent infrastructure Voice agent Delight.ai introduces Intelligent Voice Handoff, which keeps a caller connected while the agent reaches and briefs the person taking over. Answer detection distinguishes people from voicemail, while recovery handles busy lines and unanswered calls by returning to the caller, leaving a message, or creating a follow up case. Supported integrations can attach the summary, transcript, and sentiment to a ticket before the transfer.

Why it matters: Support teams can preserve the work already completed by an AI agent when a case needs human judgment. Failed transfers retain a follow up path, reducing the risk of callers being stranded between automated and human support.

Impact: Medium. Verification: Verified. Type: Workflow orchestration.

Evidence: Official blog

Source: Agentic Index change log. Free to reuse with a link back. Terms.

]]> xpander.ai added a Secrets section that declares an agent's required credentials, validates personal connections, and resolves secrets on… https://agenticindex.io/change-log/xpander-ai agenticindex.io:change:xpander-ai-1gn0f1-2026-09-20 Sun, 20 Sep 2026 12:00:00 GMT Human approval / guardrails Agent infrastructure xpander.ai added a Secrets section that declares an agent's required credentials, validates personal connections, and resolves secrets on every turn. Its September 20 release also adds skill visibility scopes, grants for individual agents, and a public skill publishing API. Gated mode now holds web fetch and search calls for approval, with gated command editing available for Claude Code, Codex, and OpenCode agents.

Why it matters: Administrators gain more explicit control over credential availability, reusable skills, and outbound web activity. Teams can apply approval policies across supported coding harnesses while keeping personal connections associated with the user.

Impact: High. Verification: Partially Verified. Type: Human approval / guardrails.

Evidence: Official changelog

Source: Agentic Index change log. Free to reuse with a link back. Terms.

]]> AutoGPT Platform v0.8.0 generates secrets per installation, refuses the previously published default encryption key, and removes legacy… https://agenticindex.io/change-log/autogpt agenticindex.io:change:autogpt-g25165-2026-09-19 Sat, 19 Sep 2026 12:00:00 GMT Security / enterprise Agent builder Agent infrastructure Multi-agent platform AutoGPT Platform v0.8.0 generates secrets per installation, refuses the previously published default encryption key, and removes legacy HS256 JWT verification. Existing installations using the defaults must migrate stored integration credentials to a new key. The release also adds scheduled expert routines, spending threshold approvals, and restrictions that let experts run only installed workflows with granted credentials.

Why it matters: Self hosted operators need to plan the credential migration before upgrading to avoid startup failures or losing access to connected services. Teams delegating recurring work gain clearer controls over cost and authorized actions.

Impact: High. Verification: Verified. Type: Security / enterprise.

Evidence: GitHub release

Source: Agentic Index change log. Free to reuse with a link back. Terms.

]]> Noma Security is available through the AWS Security Hub Extended console as its AI security partner, so enterprises can turn on AI… https://agenticindex.io/change-log/noma-security agenticindex.io:change:noma-security-zkqq6f-2026-09-18 Fri, 18 Sep 2026 12:00:00 GMT Integrations Security / SOC agent Agent infrastructure Noma Security is available through the AWS Security Hub Extended console as its AI security partner, so enterprises can turn on AI security monitoring from within AWS.

Why it matters: Buying through the console an organization already uses shortens procurement. Check which Noma capabilities are included in the Security Hub offering.

Impact: Medium. Verification: Partially Verified. Type: Integrations.

Evidence: Official blog

Source: Agentic Index change log. Free to reuse with a link back. Terms.

]]>