Jean Tessier’s Reviews > AI Agents in Action, Second Edition: Intelligent workflows with LLMs, MCP, A2A, and more > Status Update

Jean Tessier
Jean Tessier is on page 334 of 391
Putting all the pieces of the cognitive agent together. Also presents an evaluation framework to track the performance of the cognitive agent. Its power comes from combining pieces together, not from any single one being bigger or significantly better.
— Nov 01, 2026 10:17PM
AI Agents in Action, Second Edition: Intelligent workflows with LLMs, MCP, A2A, and more

flag

Jean’s Previous Updates

Jean Tessier
Jean Tessier is on page 354 of 391
Grounded tips with lots of sample agents and code snippets. It even has a few more involved examples: a customer support system, a RAG system, a research system that can search different sources. Each piece is governed by natural language description strings instead of deterministic code.
— Nov 02, 2026 11:26PM
AI Agents in Action, Second Edition: Intelligent workflows with LLMs, MCP, A2A, and more


Jean Tessier
Jean Tessier is on page 321 of 391
Multiple agents collaborate to create a cognitive agent. Almost all of them have their behavior in natural language and run by an LLM. One is more deterministic and its behavior is in code. A cognitive workspace connects them together and allows them to see what the others are doing, helping guide the overall processing.
— Oct 30, 2026 10:01PM
AI Agents in Action, Second Edition: Intelligent workflows with LLMs, MCP, A2A, and more


Jean Tessier
Jean Tessier is on page 307 of 391
Mapping failings or limitations of simpler agents to cognitive deficits. Using theory from Minsky, Baars, and Kahneman to fill in the gaps.
— Oct 28, 2026 11:13PM
AI Agents in Action, Second Edition: Intelligent workflows with LLMs, MCP, A2A, and more


Jean Tessier
Jean Tessier is on page 298 of 391
Variations on the agentic loop: adding successive layers of external control, using typed outputs to be able to combine results and determine completion. Orchestration versus collaboration (choreography) just like with microservices. The program is a loop that makes agent calls.
— Oct 26, 2026 09:19PM
AI Agents in Action, Second Edition: Intelligent workflows with LLMs, MCP, A2A, and more


Jean Tessier
Jean Tessier is on page 269 of 391
Deployment options: run in the browser, run in a service, run in a queue worker. Use Docker. Manage permissions. Build a threat model based on what you're doing and what date it uses.
— Sep 25, 2026 06:49PM
AI Agents in Action, Second Edition: Intelligent workflows with LLMs, MCP, A2A, and more


Jean Tessier
Jean Tessier is on page 232 of 391
Test-driven agent development: code agents by manipulating LLM instructions and prompts. Use agents to check the work of other agents: evaluators, grounding, critics. Critics use rubrics, elaborate prompts that tell the agent what to look for and how to rate it. Phoenix, Langfuse, or LangSmith to capture telemetry (and OpenTelemetry) and diagnose performance and errors.
— Sep 10, 2026 10:30PM
AI Agents in Action, Second Edition: Intelligent workflows with LLMs, MCP, A2A, and more


Jean Tessier
Jean Tessier is on page 196 of 391
The bit trying to relate agent "memory" to human cognitive neuroscience didn't land for me. Not sure why agent would care about "sensory memory" as opposed to image/video just being a different format of data. A lot of the integration is done in prompts, but the discussion lacks the subtleties of telling the LLM when to use what tool. (I guess that changes quickly as models evolve.)
— Sep 07, 2026 05:39PM
AI Agents in Action, Second Edition: Intelligent workflows with LLMs, MCP, A2A, and more


Jean Tessier
Jean Tessier is on page 196 of 391
RAG, search, memory: vector db are great for semantic search (similarity). Use an embedding model to converts pieces of data into vectors. Vector search usually looks at angle, not scale, only if they point in the same direction. RAG can combine from multiple sources and rerank when using hybrid search. Easy setup: ChromaDB (vector DB), server-memory MCP (graph DB). Context = short-term, external = long-term.
— Sep 07, 2026 05:32PM
AI Agents in Action, Second Edition: Intelligent workflows with LLMs, MCP, A2A, and more


Jean Tessier
Jean Tessier is on page 158 of 391
Instructing the agent how to think: chain-of-thought CoT for single pass and ReAct for iterative. Tree-of-through does breadth-first search, Reflexion does depth-first. They consume lots of tokens. Sequential Thinking MCP is a scratch pad for the agent, not thinking itself. Good examples with tools and MCP. Reflexion needs to know the answer, so it's weird. Maybe for drafting plans??
— Aug 27, 2026 10:32PM
AI Agents in Action, Second Edition: Intelligent workflows with LLMs, MCP, A2A, and more


Jean Tessier
Jean Tessier is on page 140 of 391
"An agent is a system built around an LLM, combining the model with tools, memory, orchestration logic, and the ability to interact with an environment. The LLM is the reasoning component. Everything else turns that reasoning into work that affects the world." (p. 139)
— Aug 27, 2026 09:54PM
AI Agents in Action, Second Edition: Intelligent workflows with LLMs, MCP, A2A, and more


No comments have been added yet.