Agent Capabilities for AI Agentss
Agent capabilities are the abilities that let an agent handle longer and harder tasks: planning a sequence of steps, keeping short-term and long-term memory, and coordinating several specialised agents. They determine how far an agent can go before it loses track, repeats itself or exceeds its budget.
Stage: Patterns · Part of the AI Agents roadmap · Last reviewed 19 November 2026
What to learn in Agent Capabilities
- Make an agent write and update an explicit plan during a run
- Store and retrieve long-term facts using a vector store or key-value memory
- Summarise or trim conversation history to stay inside the context window
- Split a task across two cooperating agents with defined hand-offs
- Judge when multi-agent design adds value and when it adds only overhead
Agent Capabilities interview questions
What kinds of memory does an agent use?
Short-term memory is the current context window with messages and tool results. Long-term memory is stored outside the model, in a database or vector store, and retrieved when relevant. Summaries bridge the two by compressing old context so the agent keeps key facts.
Why can multi-agent systems perform worse than a single agent?
Each hand-off adds tokens, latency and chances for misunderstanding, and agents may duplicate work or disagree. Multi-agent designs pay off when subtasks are truly separable or need different tools and instructions. Otherwise one agent with good tools is simpler and more reliable.
Where Agent Capabilities fits in the AI Agents roadmap
Agent Capabilities belongs to the Patterns stage (advanced level), which takes about 18 hours. Architectures.
Other units in this stage
Back to the full AI Agents roadmap