Platform shift: Tools AI News

Moorcheh Positions Memanto as a Fleet-Scale Layer for Agent Memory

The company’s pitch separates user-controlled memory files from the harder work of deciding what agents should retain, share and act on. Its claimed fleet-scale layer has not been independently demonstrated in the post.

Aug 28, 2026Tools

SageMaker Feature Store Adds 25-Record Batch Writes and a Record-Discovery API

The new APIs replace one-record write loops and an identifier-discovery blind spot, while leaving teams to handle partial failures and non-snapshot pagination themselves.

Aug 28, 2026Tools

Perplexity’s GitHub-Slack Agent Uses One API Call and a One-Reply Guardrail

The useful shift is not autonomous code changes: it is a constrained investigation loop that can authenticate to one repository, trace a fix, and report back without widening its Slack or GitHub scope.

Databricks Says Async Save Cut 20B-Parameter PyTorch Checkpoints to 9 Seconds on 32 H100s

The AI Runtime components pair distributed recovery with local-NVMe data caching. The remaining operational test is whether training jobs save their input position as well as their model state.

Amazon Quick Connects to fal Through MCP, Making Creative Approval Steps Reusable

The integration is designed to turn creative generation from a string of tool switches into a repeatable process, but it still requires teams to manage third-party data sharing, API access and generation costs.

OpenAI’s 10-Day WebMCP Challenge Pushes Websites to Expose Tools for AI Agents

The contest is an early test of whether sites will publish agent-facing actions instead of relying only on interfaces designed for human clicks.

Perplexity Puts Open-Weight Models Behind One API, With OpenAI and Anthropic Formats

The compatibility layer lets developers retain familiar client interfaces while Perplexity takes on deployment routing; access is still limited to a private preview.

Anthropic Puts Model-Agnostic MHS in Preview for AI-Controlled Machinery

Anthropic’s proposed interface could reduce the bespoke work of connecting agents to equipment, but broad adoption still rests on future safety work and an open-source release.

Perplexity Puts Computer Behind MCP, Opening 400+ Services to AI Clients

The endpoint turns Computer into an execution layer for compatible assistants, but clients must preserve task state and surface logins, questions and sensitive-action approvals.

Databricks Says 300M Chart-JSON Pipeline Tops Four Multimodal Baselines on Answer Accuracy

The company’s test suggests chart values can become a useful retrieval layer instead of relying solely on page-image embeddings, though its comparison measures answer correctness and includes a synthetic benchmark it built.

OpenAI Codex Client Bundles 467 Interactive Modules on a Refreshable GenUI Layer

The bundled architecture can match a conversation message to a locally supported widget and refresh its state. It still does not establish which modules are live or whether outside developers will get access.

London Surgeons Remove 11mm Brain Tumour With Live AI Anatomy Mapping

The system color-coded critical anatomy during a delicate pituitary procedure, but the reported result is one patient case within a clinical trial—not a demonstrated safety advantage over standard surgery.

Estuary Makes Rust Runtime Default, With Exactly-Once Delivery Still Conditional

The change is built to keep related database updates together as they reach agents and operational systems. But a complete end-to-end delivery guarantee still relies on the receiving destination and connector.

AWS AgentCore Evaluations Uses OpenTelemetry to Score Agents Across Frameworks

The service offers one set of agent-quality evaluators across SDK choices, but its portability depends on specific telemetry conventions, scope names and complete message logs.

SageMaker SDK v3 Lets Teams Change Model Code Without Rebuilding Container Images

The new workflow treats the container as a reusable runtime and sends local code into it when a job starts. That can shorten script-level iteration, while leaving teams responsible for images, permissions and storage.

Aug 25, 2026Tools

AWS Publishes a Phone-Ordering AI Pattern That Connects Restaurant Data to Claude

The sample shows how a voice agent can take an order from a phone call through backend tools, but restaurants must supply and operate the systems holding their own menus, carts, orders and locations.

Aug 25, 2026Tools

OpenAI’s 10-Day WebMCP Challenge Pushes Websites to Expose Tools for AI Agents

The contest is an early test of whether sites will publish agent-facing actions instead of relying only on interfaces designed for human clicks.

Aug 25, 2026Tools

Amazon OpenSearch Puts Trace and Service-Map Widgets Into MCP Agent Chats

The new OpenSearch capability turns an agent’s observability response into both an explanation and an inspectable dashboard view, but it depends on a local server, AWS credentials, and a configured OpenSearch UI workspace.