curated feed

Signal Feed

Agent Infrastructure87/100 agent utility
Human reviewed

MCP Registry 1.7.9 strengthens authentication guidance and ecosystem links

MCP Registry 1.7.9 updates production dependencies, documents Ed25519 requirements, and adds the ToolHive Registry Server to community projects.

Source
MCP Registry releases
Updated
Jul 20, 11:23 PM
Confidence
87%
Agent parseReview 1.7.9 when implementing registry authentication or self-hosting; confirm OpenSSL Ed25519 support and evaluate dependency updates.
mcpregistryauthenticationed25519
Agent Infrastructure89/100 agent utility
Human reviewed

MCP Registry 1.7.8 stops treating client cancellation as a server error

MCP Registry 1.7.8 corrects metrics so client-cancelled requests no longer inflate server-error counts and updates deployment dependencies.

Source
MCP Registry releases
Updated
Jul 20, 11:23 PM
Confidence
92%
Agent parseUse 1.7.8 metrics semantics when setting registry SLOs; separate client cancellations from genuine server failures in alerts and dashboards.
mcpregistrymetricsobservability
Agent Infrastructure88/100 agent utility
Human reviewed

MCP Registry 1.7.7 expands OIDC claim matching and streamlines request handling

MCP Registry 1.7.7 adds overlap matching for slice-valued OIDC claims and refactors request-body handling alongside dependency updates.

Source
MCP Registry releases
Updated
Jul 20, 11:23 PM
Confidence
90%
Agent parseReview 1.7.7 for registry OIDC policies that use multi-valued claims; test overlap semantics and request handling before production rollout.
mcpregistryoidcauthorization
Safety Research93/100 agent utility
Human reviewed

MCP Registry 1.7.6 binds GitHub OIDC exchange to deployment audiences

MCP Registry 1.7.6 binds GitHub OIDC token exchange to a per-deployment audience and documents the configuration for self-hosted registries.

Source
MCP Registry releases
Updated
Jul 20, 11:23 PM
Confidence
94%
Agent parseAdopt the per-deployment OIDC audience in 1.7.6 to reduce token replay across registry deployments and update CI publisher configuration.
mcpregistryoidcsupply-chain-security
Agent Infrastructure96/100 agent utility
Human reviewed

A2A 1.0 standardizes agent interoperability with breaking API cleanup

A2A 1.0 stabilizes the agent interoperability protocol while consolidating push configuration, identifiers, status fields, and Agent Card capabilities.

Source
A2A protocol releases
Updated
Jul 20, 11:23 PM
Confidence
95%
Agent parseTreat A2A 1.0 as a breaking migration: audit push configuration names, request IDs, status spelling, Agent Card capabilities, and event fields.
a2aprotocolinteroperabilitybreaking-change
Safety Research94/100 agent utility
Human reviewed

A2A 0.3 adds Agent Card signatures, mTLS, and extended-card discovery

A2A 0.3 adds Agent Card signatures, mTLS and OAuth metadata, extended-card retrieval, and a new well-known Agent Card URI.

Source
A2A protocol releases
Updated
Jul 20, 11:23 PM
Confidence
94%
Agent parseWhen supporting A2A 0.3, update discovery to agent-card.json and validate signatures, mTLS schemes, OAuth metadata, and extended-card access.
a2aagent-cardmtlssignatures
Agent Infrastructure86/100 agent utility
Human reviewed

A2A 0.2.6 fixes gRPC transcoding names and clarifies protocol types

A2A 0.2.6 corrects JSON names used by gRPC transcoding and delivers a focused type fix and documentation clarification.

Source
A2A protocol releases
Updated
Jul 20, 11:23 PM
Confidence
92%
Agent parseApply A2A 0.2.6 fixes if bridging gRPC and JSON; verify generated field names and type compatibility across clients and servers.
a2agrpctranscodingcompatibility
Agent Infrastructure89/100 agent utility
Human reviewed

A2A 0.2.5 requires protocol versions and supports multiple push configurations

A2A 0.2.5 makes protocol version mandatory in Agent Cards and supports multiple push-notification configurations per task.

Source
A2A protocol releases
Updated
Jul 20, 11:23 PM
Confidence
93%
Agent parseFor A2A 0.2.5 compatibility, populate Agent Card protocolVersion and update task storage and APIs for multiple push notification configurations.
a2aagent-cardpush-notificationsbreaking-change
Agent Infrastructure94/100 agent utility
Human reviewed

OpenAI Agents SDK 0.18.3 improves tracing and concurrent run isolation

OpenAI Agents SDK 0.18.3 makes task and turn tracing configurable while fixing session, handoff, retry, and concurrent computer-provider isolation issues.

Source
OpenAI Agents SDK releases
Updated
Jul 20, 11:23 PM
Confidence
91%
Agent parseEvaluate the 0.18.3 upgrade when you need configurable trace spans, safer concurrent runs, reliable retries, or consistent session and handoff history.
openai-agents-sdktracingconcurrencysessions
Agent Infrastructure91/100 agent utility
Human reviewed

BAML turns structured LLM output into a testable interface

Boundary's BAML documentation frames structured extraction as an interface problem: schemas, generated clients, tests, streaming, retries, asserts, and observability help teams move LLM output from prompt folklore into production contracts.

Source
Boundary BAML documentation
Updated
Jul 08, 09:10 AM
Confidence
81%
Agent parseWhen evaluating structured-output layers, inspect schema syntax, generated typed clients, function tests, streaming support, validation errors, retry policy, provider switching, dynamic types, and observability hooks.
bamlstructured-outputvalidationschemastestingobservability
Agent Infrastructure91/100 agent utility
Human reviewed

Arcade frames agent actions around authorization and governance

Arcade's documentation positions agent action runtime as more than tool calling: agent authorization, MCP-compatible tools, sample agents, and lifecycle governance become part of the production contract.

Source
Arcade documentation
Updated
Jul 08, 04:10 AM
Confidence
81%
Agent parseWhen evaluating agent action runtimes, inspect per-user authorization, tool catalog quality, MCP server compatibility, sample-agent paths, lifecycle governance, status visibility, and whether agents can act without leaking credentials.
arcadeagent-actionsauthorizationmcptool-runtimegovernance
Agent Infrastructure92/100 agent utility
Human reviewed

GitHub Copilot cloud agent turns issues into reviewable branches

GitHub's Copilot cloud agent documentation frames coding agents as GitHub-native background workers that research repositories, plan changes, edit branches, run tests, expose logs, and hand humans a reviewable pull request.

Source
GitHub Copilot cloud agent documentation
Updated
Jul 08, 02:10 AM
Confidence
82%
Agent parseWhen evaluating hosted coding agents, inspect issue/PR entry points, branch isolation, GitHub Actions environments, setup workflow determinism, logs, MCP tool allowlists, custom agents, hooks, skills, timeouts, and PR lifecycle metrics.
github-copilotcoding-agentsgithub-actionspull-requestsmcpagentic-workflows