1# Agent Design Patterns
2 
3This file covers decision heuristics for building agents on the Claude API: which primitives to reach for, how to design your tool surface, and how to manage context and cost over long runs. For per-tool mechanics and code examples, see `tool-use-concepts.md` and the language-specific folders.
4 
5---
6 
7## Model Parameters
8 
9| Parameter | When to use it | What to expect |
10| --- | --- | --- |
11| **Adaptive thinking** (`thinking: {type: "adaptive"}`) | When you want Claude to control when and how much to think. | Claude determines thinking depth per request and automatically interleaves thinking between tool calls. No token budget to tune. |
12| **Effort** (`output_config: {effort: ...}`) | When adjusting the tradeoff between thoroughness and token efficiency. | Lower effort → fewer and more-consolidated tool calls, less preamble, terser confirmations. `medium` is often a favorable balance. Use `max` when correctness matters more than cost. |
13 
14See `SKILL.md` §Thinking & Effort for model support and parameter details.
15 
16---
17 
18## Designing Your Tool Surface
19 
20### Bash vs. dedicated tools
21 
22Claude doesn't know your application's security boundary, approval policy, or UX surface. Claude emits tool calls; your harness handles them. The shape of those tool calls determines what the harness can do.
23 
24A **bash tool** gives Claude broad programmatic leverage — it can perform almost any action. But it gives the harness only an opaque command string, the same shape for every action. Promoting an action to a **dedicated tool** gives the harness an action-specific hook with typed arguments it can intercept, gate, render, or audit.
25 
26**When to promote an action to a dedicated tool:**
27 
28- **Security boundary.** Actions that require gating are natural candidates. Reversibility is a useful criterion: hard-to-reverse actions (external API calls, sending messages, deleting data) can be gated behind user confirmation. A `send_email` tool is easy to gate; `bash -c "curl -X POST ..."` is not.
29- **Staleness checks.** A dedicated `edit` tool can reject writes if the file changed since Claude last read it. Bash can't enforce that invariant.
30- **Rendering.** Some actions benefit from custom UI. Claude Code promotes question-asking to a tool so it can render as a modal, present options, and block the agent loop until answered.
31- **Scheduling.** Read-only tools like `glob` and `grep` can be marked parallel-safe. When the same actions run through bash, the harness can't tell a parallel-safe `grep` from a parallel-unsafe `git push`, so it must serialize.
32 
33**Rule of thumb:** Start with bash for breadth. Promote to dedicated tools when you need to gate, render, audit, or parallelize the action.
34 
35---
36 
37## Anthropic-Provided Tools
38 
39| Tool | Side | When to use it | What to expect |
40| --- | --- | --- | --- |
41| **Bash** | Client | Claude needs to execute shell commands. | Claude emits commands; your harness executes them. Reference implementation provided. |
42| **Text editor** | Client | Claude needs to read or edit files. | Claude views, creates, and edits files via your implementation. Reference implementation provided. |
43| **Computer use** | Client or Server | Claude needs to interact with GUIs, web apps, or visual interfaces. | Claude takes screenshots and issues mouse/keyboard commands. Can be self-hosted (you run the environment) or Anthropic-hosted. |
44| **Code execution** | Server | Claude needs to run code in a sandbox you don't want to manage. | Anthropic-hosted container with built-in file and bash sub-tools. No client-side execution. |
45| **Web search / fetch** | Server | Claude needs information past its training cutoff (news, current events, recent docs) or the content of a specific URL. | Claude issues a query or URL; Anthropic executes it and returns results with citations. |
46| **Memory** | Client | Claude needs to save context across sessions. | Claude reads/writes a `/memories` directory. You implement the storage backend. |
47 
48**Client-side** tools are defined by Anthropic (name, schema, Claude's usage pattern) but executed by your harness. Anthropic provides reference implementations. **Server-side** tools run entirely on Anthropic infrastructure — declare them in `tools` and Claude handles the rest.
49 
50---
51 
52## Composing Tool Calls: Programmatic Tool Calling
53 
54With standard tool use, each tool call is a round trip: Claude calls the tool, the result lands in Claude's context, Claude reasons about it, then calls the next tool. Three sequential actions (read profile → look up orders → check inventory) means three round trips. Each adds latency and tokens, and most of the intermediate data is never needed again.
55 
56**Programmatic tool calling (PTC)** lets Claude compose those calls into a script instead. The script runs in the code execution container. When the script calls a tool, the container pauses, the call is executed (client-side or server-side), and the result returns to the running code — not to Claude's context. The script processes it with normal control flow (loops, filters, branches). Only the script's final output returns to Claude.
57 
58| When to use it | What to expect |
59| --- | --- |
60| Many sequential tool calls, or large intermediate results you want filtered before they hit the context window. | Claude writes code that invokes tools as functions. Runs in the code execution container. Token cost scales with final output, not intermediate results. |
61 
62---
63 
64## Scaling the Tool and Instruction Set
65 
66| Feature | When to use it | What to expect |
67| --- | --- | --- |
68| **Tool search** | Many tools available, but only a few relevant per request. Don't want all schemas in context upfront. | Claude searches the tool set and loads only relevant schemas. Tool definitions are appended, not swapped — preserves cache (see Caching below). |
69| **Skills** | Task-specific instructions Claude should load only when relevant. | Each skill is a folder with a `SKILL.md`. The skill's description sits in context by default; Claude reads the full file when the task calls for it. |
70 
71Both patterns keep the fixed context small and load detail on demand.
72 
73---
74 
75## Long-Running Agents: Managing Context
76 
77| Pattern | When to use it | What to expect |
78| --- | --- | --- |
79| **Context editing** | Context grows stale over many turns (old tool results, completed thinking). | Tool results and thinking blocks are cleared based on configurable thresholds. Keeps the transcript lean without summarizing. |
80| **Compaction** | Conversation likely to reach or exceed the context window limit. | Earlier context is summarized into a compaction block server-side. See `SKILL.md` §Compaction for the critical `response.content` handling. |
81| **Memory** | State must persist across sessions (not just within one conversation). | Claude reads/writes files in a memory directory. Survives process restarts. |
82 
83**Choosing between them:** Context editing and compaction operate within a session — editing prunes stale turns, compaction summarizes when you're near the limit. Memory is for cross-session persistence. Many long-running agents use all three.
84 
85---
86 
87## Caching for Agents
88 
89**Read `prompt-caching.md` first.** It covers the prefix-match invariant, breakpoint placement, the silent-invalidator audit, and why changing tools or models mid-session breaks the cache. This section covers only the agent-specific workarounds for those constraints.
90 
91| Constraint (from `prompt-caching.md`) | Agent-specific workaround |
92| --- | --- |
93| Editing the system prompt mid-session invalidates the cache. | Append a `<system-reminder>` block in the `messages` array instead. The cached prefix stays intact. Claude Code uses this for time updates and mode transitions. |
94| Switching models mid-session invalidates the cache. | Spawn a **subagent** with the cheaper model for the sub-task; keep the main loop on one model. Claude Code's Explore subagents use Haiku this way. |
95| Adding/removing tools mid-session invalidates the cache. | Use **tool search** for dynamic discovery — it appends tool schemas rather than swapping them, so the existing prefix is preserved. |
96 
97For multi-turn breakpoint placement, use top-level auto-caching — see `prompt-caching.md` §Placement patterns.
98 
99---
100 
101For live documentation on any of these features, see `live-sources.md`.
102