Building, evaluating, and running AI agents: frameworks, development environments, memory, tools, MCP, Codex and Claude Code in practice.
Start with the three meanings of agent development environment: a coding product, an RL task system, or an execution runtime. Then inspect the RL environment projects or keep the short definition handy.
A focused reading order to help you choose your next step.
Separate the three meanings of agent development environment: coding-agent products, RL training environments and hosted execution runtimes.
A practical research-agent design: preserve sources, separate facts from inferences, and deliver a decision someone can review.
Learn what Model Context Protocol (MCP) is, how it works, and why it's the universal standard connecting AI agents to tools and data.
Step-by-step tutorial for building a multi-agent AI system from scratch using CrewAI, LangGraph, or AutoGen — with architecture patterns and production tips.
A 7-step engineer's guide to deploy AI agents production-ready in 2026: hosting, state, observability, evals, retries, cost controls, and rollouts.
Learn how to monitor and debug AI agents with traces, metrics, alerts, and replay evals. Stop guessing why your agent failed in production.
Build a small ticket-label agent in stages: define its boundaries, add retrieval and state, then test failure and recovery. Each lesson has an exercise and a concrete deliverable.
Download the local lab. The downloadable Python 3.10+ lab uses only the standard library and synthetic data. Run python3 agent_lab.py for eight boundary tests. It makes no network or model calls and does not establish production readiness.
Send a request, parse its response and test API failures.
Verify a delivery before accepting its work.
Write three prompts with explicit output contracts.
Separate a model proposal from an authorized action.
Run an MCP server and client; discover and test a tool.
Filter retrieval by access and test insufficient evidence.
Scope and expire remembered facts.
Bound a plan and stop when a dependency fails.
Bind approval to the exact proposed change.
Run a dataset and preserve failures in explicit scoring denominators.
Reproduce a local evaluation and compare tools on the same task.
Run four control flows and inspect their failure behavior.
Audit the hidden assumptions in a starter.
Review a constrained coding change against behavior.
Make one small app change with a testable contract.
Import and execute a ticket-triage workflow with failure fixtures.
Connect signed delivery, retrieval and approval; verify restart and replay.
Run a fixed workflow, router, bounded tool loop and parallel join. Inspect which steps run, which stop and what evidence reaches the reviewer.
Published
Compare a Python agent backend, a LangGraph chat UI and a full-stack chatbot, then reproduce a pinned starter inspection without paid inference.
Published · Updated
Reproduce a local Pydantic Evals run, then compare Promptfoo, LangSmith, Braintrust and plain Python against the same dataset and scoring rules.
Published · Updated
Make a Claude Messages API request in Python, validate its JSON response, and test authentication, rate limits and bad output with local fixtures.
Published
Run ten ticket cases through two deterministic baselines, inspect wrong labels and timeouts, and retain failures in a reproducible evaluation report.
Published · Updated
Run a SQLite approval lifecycle that rejects changed, stale, expired and revoked proposals, then returns a completed local receipt on retry.
Published · Updated
Run a four-step ticket workflow with validated dependencies, a dispatch budget, cancellation and traces of failed prerequisites.
Published · Updated
Run a SQLite preference store with owner scoping, versioned updates, provenance, expiry and deletion, then verify it from fresh Python processes.
Published · Updated
Run a six-document retrieval and answer pipeline, inspect citations, exclude private and outdated evidence, and measure an intentional synonym failure.
Published · Updated
Run a versioned Python MCP server and client, discover a ticket-label tool, and test valid calls, schema failures and protocol errors.
Published · Updated
Run a Python tool-calling loop with schemas, argument checks, matched results, turn limits and ten deterministic success and failure fixtures.
Published · Updated
Build one local ticket-review flow in Cursor, then verify proposal, approval, stale rejection and retry behavior through the UI and real HTTP checks.
Published · Updated
Use Copilot in VS Code on a small SQLite bug, starting with two failing tests and reviewing the exact version check in the resulting patch.
Published · Updated
Run a local Python webhook receiver, verify signed bytes, persist event receipts, and test duplicate delivery, tampering and worker rollback.
Published
Build classification, grounded-answer and change-proposal prompts with explicit outputs, twelve fixtures, and a runnable Python evaluator.
Published
Compare five RL environment and task-generation projects for coding agents, with primary sources, selection criteria and a CSV export.
Published
Connect a real HTTP receiver to retrieval, proposal validation and a stored approval, then verify twelve local failure and recovery scenarios.
Published
Reactive vs proactive AI agents explained: architecture, triggers, planning loops, risks, examples, and when to use each pattern.
Published · Updated
A practical research-agent design: preserve sources, separate facts from inferences, and deliver a decision someone can review.
Published · Updated
Separate the three meanings of agent development environment: coding-agent products, RL training environments and hosted execution runtimes.
Published · Updated
Compare five enterprise agent deployment options by identity, state, integration and operations, then run a pilot with explicit acceptance criteria.
Published · Updated
A source-based shortlist of Datalab, ABBYY, Rossum, Google, and Azure, with practical questions about extraction, review, and deployment.
Published · Updated
Compare Glean, Guru, Notion and Microsoft knowledge search through connector permissions, freshness, answer evidence and a concrete customer pilot.
Published · Updated
Learn how to monitor and debug AI agents with traces, metrics, alerts, and replay evals. Stop guessing why your agent failed in production.
Published · Updated
A 7-step engineer's guide to deploy AI agents production-ready in 2026: hosting, state, observability, evals, retries, cost controls, and rollouts.
Published · Updated
Learn what Model Context Protocol (MCP) is, how it works, and why it's the universal standard connecting AI agents to tools and data.
Published · Updated
Step-by-step tutorial for building a multi-agent AI system from scratch using CrewAI, LangGraph, or AutoGen — with architecture patterns and production tips.
Published · Updated