Skip to content
AI.info

AI.info

AI agents

Action under control: goals, tools, permissions, stop conditions, and who answers for what the system did.

  1. Agentic Systems as Controlled Action
  2. Workflow, Assistant, or Agent?
  3. Goals, Task Contracts, and Success Criteria
  4. Autonomy, Authority, and Delegation
  5. Environment, Observation, State, and Action
  6. Agent Loops, Harnesses, and Runtimes
  7. Deterministic Control and Model Judgment
  8. Architecture Decision Map for Agentic Systems
  9. Function Calling and Typed Action Contracts
  10. Tool Schema Design and Agent Affordances
  11. Tool Discovery, Selection, and Routing
  12. Tool Results, Provenance, and State Updates
  13. Timeouts, Retries, and Idempotent Tool Use
  14. Transactions, Side Effects, and Compensation
  15. Credentials, Scopes, and Least Privilege
  16. Search and Retrieval Tools for Agents
  17. Browser Agents and Web Navigation
  18. Code Execution and Sandboxed Compute
  19. Computer Use and GUI Control
  20. Model Context Protocol: Tools, Resources, and Prompts
  21. Agent2Agent Protocol and Remote Agent Collaboration
  22. Tool Contract Testing and Tool-Use Evaluation
  23. ReAct and the Action–Observation Loop
  24. Task Decomposition and Dependency Graphs
  25. Plan-and-Execute Versus Stepwise Control
  26. Replanning, Branching, and Search Budgets
  27. Verifiers, Tests, Critics, and Outcome Checks
  28. Reflection and Self-Correction Without Magical Thinking
  29. State Machines, Policies, and Explicit Orchestration
  30. Parallelism, Concurrency, and Synchronization
  31. Clarification, Uncertainty, and Abstention
  32. Long-Horizon Work, Checkpoints, and Resumability
  33. Termination, Loop Detection, and Safe Stopping
  34. Context Engineering and Attention Budgets
  35. Working Memory, Scratch State, and Task Ledgers
  36. Episodic Memory and Event Histories
  37. Semantic Memory and Knowledge Stores
  38. Memory Write Policies, Retrieval, and Forgetting
  39. Summarization, Compaction, and Context Drift
  40. Durable State, Provenance, Replay, and Recovery
  41. Identity, Privacy, Retention, and Deletion
  42. When Multi-Agent Systems Help—or Hurt
  43. Supervisor–Worker Orchestration
  44. Handoffs, Specialists, and Capability Routing
  45. Parallel Research and Map–Reduce Patterns
  46. Critique, Debate, and Result Aggregation
  47. Shared State, Messaging, and Coordination
  48. Deadlocks, Cascading Errors, and Conflict Resolution
  49. Human–Agent Teams, Escalation, and Mixed Initiative
  50. Research Agents and Evidence Synthesis
  51. Coding Agents and Repository Workflows
  52. Transactional and Customer-Service Agents
  53. Agent Evaluation Starts With Environment State
  54. Trajectory, Tool, and Policy Evaluation
  55. Reliability, Simulation, and Benchmark Design
  56. Observability, Tracing, and Systematic Agent Debugging
  57. Prompt Injection, Secrets, and the Confused Deputy
  58. Capability Control, Approvals, Sandboxing, and Rollback
  59. Cost, Latency, Deployment, Monitoring, and Incidents
  60. AI Agents Capstone: Design, Evaluate, and Govern an Agentic System