The August 10 field edition

Before you trust an agent, check the source.

The latest releases changed how agents handle tools, data, sessions, approvals, and handoffs. We traced each change to the official source. Start with five recent changes before your next build.

A physical miniature agent system links memory, tools, approval, artifact handling and handoff around a blue task route.
One task moves through memory, tools, approval, artifacts, and handoff. The blue route marks every place the work can wait, change, or pass to someone else. AI-assisted editorial illustration.

In this editionA live index of what you can read today.

656
sourced records
301
systems and frameworks
157
terms and concepts
2
records held for more proof

The week in the field

Which five changes should you check before your next build?

These releases changed runtime context, data handling, approvals, sessions, or handoffs. Each note starts with the official release. Find the change closest to your work, then read the source before you ship. That is the quickest way to see whether it changes your limits, tests, or review plan.

A blue route crosses three square checkpoints before meeting a black mechanical stop in a miniature architectural model.
A good trace shows the route, every checkpoint, and the place a person stopped the work. AI-assisted editorial illustration.

Deep Agents separates private subagent state

LangChain QuickJS 0.3.7 keeps private state out of subagent propagation while adding native streaming for programmatic tool calls. The preceding 0.3.6 release added an HMAC check for serialized QuickJS snapshots; neither change makes arbitrary generated code safe outside the runtime's other isolation and policy controls.

Read the primary source
A cutaway architectural model shows one blue path moving through guarded black-and-bone chambers.
A file boundary has to hold from the first check to the place where the session writes. AI-assisted editorial illustration.

Pydantic AI makes realtime interruption state explicit

Pydantic AI 2.40 adds barge-in handling, caller-reported played-audio byte counts, out-of-band prompts, and agent event listeners. Accurate rollback depends on the driver reporting how much buffered audio actually played; the API cannot infer external playback by itself.

Read the source
Blue tokens move through transparent and black test frames while one route diverts into a safe holding tray.
The valid path continues. The rejected action waits where someone can inspect it. AI-assisted editorial illustration.

CrewAI closes its internal model-hook gap

CrewAI 1.15.19 runs model-call hooks on every execution path and propagates deny decisions, superseding the 1.15.18 limitation for CrewAI-internal flows. The same release bumps named document-processing dependencies for security vulnerabilities, but its notes do not identify CVEs for those two bumps.

Read the source
Black-and-bone tool modules circle a selector that feeds one chosen module onto a blue execution rail.
A skill can be picked and reused like a tool. Someone still has to own its permissions and release. AI-assisted editorial illustration.

Agno negotiates stateless MCP explicitly

Agno 3.0.6 can serve MCP without session tracking and lets clients choose legacy, automatic, or current protocol behavior. Stateless mode stays off by default and gives up server-initiated notifications and resumable SSE, so horizontal routing improves only by accepting those feature losses.

Read the source
A blue command route enters a guarded black-and-bone artifact chamber through layered inspection gates.
The tool call crosses a confirmation gate before it can touch the artifact. AI-assisted editorial illustration.

Mastra restores context to durable approval checks

Mastra core 1.64 passes request context and workspace data into per-tool approval functions on durable and networked agents, including after cross-process resume. The release also validates reconstructed MCP structured output and warns when authenticated servers trust client-supplied resource identifiers without a user-to-resource mapping.

Read the source

The reference desks

What do you need to know right now?

Start with the question in front of you. Need a definition? Open the glossary. Choosing a framework? Compare systems. Checking who built what? Use the ecosystem. Looking for a rule? Read the standards. One clear question is enough to begin.

Start here · 01–05

Which five pages make the rest easier?

They give you the agent, its workflow, its limits, its tool protocol, and the evaluation that helps you judge the result. Read them in order and the rest of the field becomes easier to follow. Together, they give you a working frame for every system, standard, and release note that follows.

  1. 01AI agent
  2. 02Agent workflow
  3. 03Agent governance
  4. 04Model Context Protocol
  5. 05Agent evaluation

The visual report

Which record should you read next?

Start with any term, system, organization, or standard. The map shows the records sitting closest to it, then lets you open each source for yourself. Use those connections to choose what to read next. A line suggests a path; it does not prove that the two things belong together.

Explore the full field map
Drawing the field plate

The map will appear here when its records are ready.

From the field

Found a problem worth fixing?

Turwin builds the system when the answer needs more than a definition.

Keep the guide

Want a copy you can keep?

Get the glossary with its dates and sources.

Send this form only if you want a reply about the guide or your request. We do not sell the list, and a person decides how we respond.

Learn · Explore · Build

Where should you go next?

  1. 01

    AgentMode.ai

    Learn the agent field.

  2. 02

    TurwinLabs.com

    Explore ideas worth testing.

  3. 03

    Turwin.ai

    Build the venture.