1# Platform Availability
2 
3Which features work on which provider platform. **This table is the single source of truth in this skill** — per-feature sections elsewhere point here instead of restating availability. When writing code for a third-party platform (Bedrock, Vertex, Foundry) or Claude Platform on AWS, check this table first; a feature not supported there means use the first-party Claude API surface or a different approach.
4 
5Columns: **1P** = first-party Claude API, **P-AWS** = Claude Platform on AWS (Anthropic-operated, same-day parity), **Bedrock** = Amazon Bedrock, **Vertex** = Google Cloud Vertex AI, **Foundry** = Microsoft Foundry. ✅ = GA, β = beta, ❌ = not supported.
6 
7| Feature | 1P | P-AWS | Bedrock | Vertex | Foundry | Notes |
8|---|---|---|---|---|---|---|
9| Messages, streaming, tool use | ✅ | ✅ | ✅ | ✅ | ✅ | Core API |
10| PDF input | ✅ | ✅ | ✅ | ✅ | β | |
11| Structured outputs / strict tool use | ✅ | ✅ | ✅ | ✅ | β | |
12| Adaptive thinking / effort | ✅ | ✅ | ✅ | ✅ | β | |
13| Extended thinking | ✅ | ✅ | ✅ | ✅ | β | |
14| Prompt caching (5m, 1h) | ✅ | ✅ | ✅ | ✅ | β | |
15| Automatic prompt caching | ✅ | ✅ | ❌ | ❌ | β | |
16| Token counting | ✅ | ✅ | ✅ | ✅ | β | |
17| Citations | ✅ | ✅ | ✅ | ✅ | β | |
18| Search results content blocks | ✅ | ✅ | ✅ | ✅ | β | |
19| Fine-grained tool streaming | ✅ | ✅ | ✅ | ✅ | ✅ | |
20| Compaction | β | β | β | β | β | |
21| Context editing | β | β | β | β | β | |
22| Context windows (1M) | ✅ | ✅ | ✅ | ✅ | β | |
23| `inference_geo` (data residency) | ✅ | ✅ | ❌ | ❌ | ❌ | |
24| **Server-side tools** | | | | | | |
25|   Web search | ✅ | ✅ | ❌ | ✅ | β | Vertex: basic `web_search_20250305` only (no `_20260209` dynamic filtering) |
26|   Web fetch | ✅ | ✅ | ❌ | ❌ | β | |
27|   Code execution | ✅ | ✅ | ❌ | ❌ | β | |
28|   Tool search | ✅ | ✅ | ✅ | ✅ | β | Bedrock: InvokeModel API only, not Converse |
29|   Advisor tool | β | β | ❌ | ❌ | ❌ | |
30| **Client-implemented tools** | | | | | | |
31|   Bash, text editor, memory | ✅ | ✅ | ✅ | ✅ | β | |
32|   Computer use | β | β | β | β | β | |
33| **Agentic / orchestration** | | | | | | |
34|   Agent Skills (Messages API) | β | β | ❌ | ❌ | β | |
35|   Programmatic tool calling | ✅ | ✅ | ❌ | ❌ | β | |
36|   MCP connector | β | β | ❌ | ❌ | β | |
37|   Managed Agents | β | β | ❌ | ❌ | ❌ | Foundry ❌ inferred (not in Foundry docs either way) |
38|   Self-hosted sandboxes | β | β | ❌ | ❌ | ❌ | P-AWS: `GET /v1/environments/{id}/work` list endpoint not supported; other work endpoints OK |
39| **API endpoints** | | | | | | |
40|   Message Batches | ✅ | ✅ | ❌ | ❌ | ❌ | |
41|   Files API | β | β | ❌ | ❌ | β | |
42|   Models API | ✅ | ✅ | ❌ | ❌ | ❌ | |
43| **Other** | | | | | | |
44|   Mid-conversation system messages | ✅ | ✅ | ❌ | ❌ | ❌ | Claude Opus 4.8 only |
45|   Fast mode | β | ❌ | ❌ | ❌ | ❌ | Research preview, beta `fast-mode-2026-02-01`, first-party API only |
46|   Cache diagnostics | β | ❌ | ❌ | ❌ | ❌ | First-party API only |
47|   Task budgets | β | β | ❌ | ❌ | ❌ | Beta header `task-budgets-2026-03-13`; 3P availability not documented — assume unsupported |
48 
49<!--
50GROUNDING (reviewer-only; stripped at runtime by processSkillMarkdown).
51All paths are under docker_eval/resources/cdp-skill/public-docs/.
52 
53Primary source: build-with-claude/overview.mdx <PlatformAvailability> props
54(claudeApi→1P, claudePlatformAws→P-AWS, bedrock→Bedrock, vertexAi→Vertex,
55azureAi→Foundry; *Beta suffix→β; prop absent→❌). Per-row citations:
56 
57  Context windows          ov:44
58  Adaptive thinking        ov:45
59  Batch / Message Batches  ov:46; bed:360; vtx:381; fdy:507
60  Citations                ov:47
61  inference_geo            ov:48
62  Effort                   ov:49
63  Extended thinking        ov:50
64  PDF input                ov:51
65  Search results           ov:52
66  Structured outputs       ov:53
67  Advisor tool             ov:63
68  Code execution           ov:64
69  Web fetch                ov:65
70  Web search               ov:66; agents-and-tools/tool-use/web-search-tool.mdx:41
71  Bash/text-editor/memory  ov:72,75,74
72  Computer use             ov:73
73  Agent Skills             ov:83
74  Fine-grained streaming   ov:84
75  MCP connector            ov:85; agents-and-tools/mcp-connector.mdx:36
76  Programmatic tool call   ov:86
77  Tool search              ov:87; agents-and-tools/tool-use/tool-search-tool.mdx:24-30
78  Compaction               ov:95
79  Context editing          ov:96
80  Automatic caching        ov:97
81  Prompt caching 5m/1h     ov:98,99
82  Token counting           ov:100
83  Files API                ov:108; build-with-claude/files.mdx:17
84  Managed Agents           managed-agents/overview.mdx:11,70-72; bed:360; vtx:381
85  Self-hosted sandboxes    build-with-claude/claude-platform-on-aws.mdx:525,547
86  Mid-convo system msgs    build-with-claude/mid-conversation-system-messages.mdx:15
87  Fast mode                build-with-claude/fast-mode.mdx:23
88  Cache diagnostics        build-with-claude/cache-diagnostics.mdx:15,1379
89  Task budgets             build-with-claude/task-budgets.mdx:15
90  Models API               bed:360; vtx:381; fdy:506
91 
92  ov  = build-with-claude/overview.mdx
93  bed = build-with-claude/claude-in-amazon-bedrock.mdx
94  vtx = build-with-claude/claude-on-vertex-ai.mdx
95  fdy = build-with-claude/claude-in-microsoft-foundry.mdx
96-->
97