Skip to content

fix: suppress output schema for ToolResult subclass annotations#3548

Merged
jlowin merged 3 commits intomainfrom
fix/tool-result-subclass-schema
Mar 18, 2026
Merged

fix: suppress output schema for ToolResult subclass annotations#3548
jlowin merged 3 commits intomainfrom
fix/tool-result-subclass-schema

Conversation

@jlowin
Copy link
Copy Markdown
Member

@jlowin jlowin commented Mar 18, 2026

replace_type uses exact type identity to recognize ToolResult and suppress automatic output schema generation — but subclasses slip through. A tool annotated -> MyToolResult(ToolResult) gets a spurious schema generated from the container's internal Pydantic fields (content, structured_content, etc.), which then fails MCP SDK validation at runtime because the actual structured content doesn't match the container's shape.

The fix adds _contains_tool_result_type(), mirroring the existing _contains_prefab_type() pattern — a subclass-aware check that recurses through unions and Annotated, applied before replace_type runs.

class MyToolResult(ToolResult):
    def __init__(self, data: str):
        super().__init__(structured_content={"content": data})

@server.tool()
async def my_tool() -> MyToolResult:  # no longer generates a bogus schema
    return MyToolResult("hello")

Closes #3528

@jlowin jlowin added the bug Something isn't working. Reports of errors, unexpected behavior, or broken functionality. label Mar 18, 2026
@marvin-context-protocol marvin-context-protocol Bot added the server Related to FastMCP server implementation or server-side functionality. label Mar 18, 2026
Copy link
Copy Markdown

@chatgpt-codex-connector chatgpt-codex-connector Bot left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: b91d5950cd

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

  • Open a pull request for review
  • Mark a draft as ready
  • Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

Comment on lines +225 to +228
# ToolResult subclasses should suppress schema generation just
# like ToolResult itself — replace_type only does exact matching.
if _contains_tool_result_type(output_type):
output_type = _UnserializableType
Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

P2 Badge Treat ToolResult subclasses as no-schema in transform fallback

When this branch forces ParsedFunction.output_schema to None for a ToolResult subclass, Tool.from_tool(..., transform_fn=...) now takes the output_schema is None fallback in src/fastmcp/tools/tool_transform.py:493-506 and inherits the parent tool's schema unless the annotation is exactly ToolResult. In the common case where the parent has an output schema and the custom transform returns a ToolResult subclass with different structured_content, the transformed tool will advertise the parent schema even though run() returns the subclass payload, which can still fail MCP output validation at runtime.

Useful? React with 👍 / 👎.

@marvin-context-protocol
Copy link
Copy Markdown
Contributor

Test Failure Analysis

Summary: The Tests with lowest-direct dependencies job fails in — this failure is unrelated to this PR's changes and is a pre-existing issue already being fixed in PR #3549.

Root Cause: The test at tests/utilities/test_components.py:240 asserts len(w) == 1 — expecting exactly one warning when creating a FastMCPComponent subclass without KEY_PREFIX. In the lowest-direct-dependency environment (pydantic==2.11.7), Pydantic emits 4 additional deprecation warnings when subclassing a BaseModel, pushing the count to 5 and failing the assertion. The PR's changes (only touching function_parsing.py and test_output_schema.py) have no effect on this test.

Suggested Solution: This is being fixed in PR #3549 (fix/flaky-key-prefix-warning-test), which filters the captured warnings by message content rather than checking the total count. Once that PR merges, this failure will be resolved.

Detailed Analysis

Failing test:

# tests/utilities/test_components.py:232
def test_warning_for_missing_key_prefix(self):
    with warnings.catch_warnings(record=True) as w:
        warnings.simplefilter("always")
        class NoPrefix(FastMCPComponent):
            pass
        assert len(w) == 1  # fails: 5 warnings emitted with pydantic 2.11.7

Log excerpt:

E           assert 5 == 1
FAILED tests/utilities/test_components.py::TestKeyPrefix::test_warning_for_missing_key_prefix
1 failed, 5003 passed, 1 skipped in 125.87s

Fix from PR #3549:

key_prefix_warnings = [
    x for x in w if "does not define KEY_PREFIX" in str(x.message)
]
assert len(key_prefix_warnings) == 1

The same failure is reproducible across multiple unrelated PRs (e.g., #3549 was opened specifically to fix it), confirming this is not a regression from this PR.

Related Files
  • tests/utilities/test_components.py — contains the failing test (TestKeyPrefix::test_warning_for_missing_key_prefix)
  • src/fastmcp/utilities/components.py — defines FastMCPComponent.__init_subclass__ which emits the KEY_PREFIX warning
  • PR fix: flaky KEY_PREFIX warning test in lowest-direct deps #3549 — fixes the flaky test by filtering warnings by message instead of checking total count

🤖 Analysis by marvin — the FastMCP triage bot

Copy link
Copy Markdown

@chatgpt-codex-connector chatgpt-codex-connector Bot left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 2eae136da7

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

  • Open a pull request for review
  • Mark a draft as ready
  • Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

Comment thread src/fastmcp/tools/tool_transform.py Outdated
Comment on lines +501 to +504
return_annotation = inspect.signature(
transform_fn
).return_annotation
if return_annotation is ToolResult:
if issubclass_safe(return_annotation, ToolResult):
Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

P2 Badge Resolve postponed return annotations before parent-schema fallback

Fresh evidence: with from __future__ import annotations (which this repo already exercises in tests/tools/test_tool_future_annotations.py), inspect.signature(transform_fn).return_annotation here is the string 'MyToolResult', so issubclass_safe(...) returns false even though ParsedFunction.from_function() has already resolved the return type and set output_schema=None. In that case Tool.from_tool(..., transform_fn=...) falls back to the parent tool's schema, and a custom transform that returns a ToolResult subclass can still advertise the wrong output schema and fail MCP output validation at runtime.

Useful? React with 👍 / 👎.

@jlowin jlowin merged commit 9aa31d5 into main Mar 18, 2026
7 checks passed
@jlowin jlowin deleted the fix/tool-result-subclass-schema branch March 18, 2026 19:22
jlowin added a commit that referenced this pull request Mar 30, 2026
Co-authored-by: Claude Opus 4.6 <[email protected]>
Co-authored-by: Jeremiah Lowin <[email protected]>
Co-authored-by: Marvin Context Protocol <41898282+Marvin Context [email protected]>
Co-authored-by: voidborne-d <[email protected]>
Co-authored-by: marvin-context-protocol[bot] <225465937+marvin-context-protocol[bot]@users.noreply.github.com>
Co-authored-by: Claude <[email protected]>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
Co-authored-by: d 🔹 <[email protected]>
Co-authored-by: Jeremiah Lowin <[email protected]>
Co-authored-by: nightcityblade <[email protected]>
Co-authored-by: Claude Opus 4.6 (1M context) <[email protected]>
Co-authored-by: Claude Sonnet 4.6 <[email protected]>
Co-authored-by: Bill Easton <[email protected]>
Co-authored-by: Sumanshu Nankana <[email protected]>
Co-authored-by: Eric Robinson <[email protected]>
Co-authored-by: Martim Santos <[email protected]>
Co-authored-by: d 🔹 <[email protected]>
Co-authored-by: Matthieu B <[email protected]>
Co-authored-by: Sascha Buehrle <[email protected]>
Co-authored-by: Hakancan <[email protected]>
Co-authored-by: nightcityblade <[email protected]>
Co-authored-by: Matt Hallowell <[email protected]>
Co-authored-by: nate nowack <[email protected]>
Co-authored-by: Bill Easton <[email protected]>
Co-authored-by: Marcus Shu <[email protected]>
Co-authored-by: Rushabh Doshi <[email protected]>
Co-authored-by: AIKAWA Shigechika <[email protected]>
Co-authored-by: Jeremy Simon <[email protected]>
Co-authored-by: Miguel Miranda Dias <[email protected]>
Co-authored-by: Anthony James Padavano <[email protected]>
Co-authored-by: Mostafa Kamal <[email protected]>
Fix auto-close MRE script posting comment without closing (#3386)
Fix WorkOS token scope verification bypass 🤖 Generated with Codex (#3407)
Fix initialize McpError fallthrough 🤖 Generated with Codex (#3413)
Fix transform arg collisions with passthrough params (#3431)
Fix get_* returning None when latest version is disabled (#3439)
Fix get_* returning None when latest version is disabled (#3421)
Fix server lifespan overlap teardown (#3415)
Fix $ref output schema object detection regression (#3420)
resolved annotations (#3429)
Fix async partial callables rejected by iscoroutinefunction (#3438)
Fix async partial callables rejected by iscoroutinefunction (#3423)
fix: add version to components (#3458)
fix: use intent-based flag for OIDC scope patch in load_access_token (#3465)
Fixes #3461
fix: normalize Google scope shorthands and surface valid_scopes (#3477)
fix: resolve ty 0.0.23 type-checking errors and bump pin (#3481)
fix: shield lifespan teardown from cancellation (#3480)
fix: forward custom_route endpoints from mounted servers (#3462)
fix updates _get_additional_http_routes() to traverse providers,
Fixes #3457
fix: remove hardcoded version from CLI help text (#3456)
fix: monty 0.0.8 compatibility, drop external_functions from constructor (#3468)
fix: task test teardown hanging 5s per test (#3499)
Closes #3498
fix: validate workspace path is a directory before cursor install (#3440)
Fixes #3426
fix: handle re.error from malformed URI templates in build_regex (#3501)
fix: reject empty/OIDC-only required_scopes in AzureProvider (#3503)
fix: restrict $ref resolution to local refs only (SSRF/LFI) (#3502)
fix warnings and timeouts (#3504)
close upgrade check issue when build passes (#3505)
Closes #3484
fix: URL-encode path params to prevent SSRF/path traversal (GHSA-vv7q-7jx5-f767) (#3507)
fix: prevent path traversal in skill download (#3493)
fix: prefer IdP-granted scopes over client-requested scopes in OAuthProxy (#3492)
fix: remove unrelated transform and http.py changes from PR scope
fix: remove forced follow_redirects from httpx_client_factory calls (#3496)
fix: stop passing follow_redirects to httpx_client_factory
fix: restore follow_redirects=True for custom httpx client factories
Closes #3509
fix: CSRF double-submit cookie check in consent flow (#3519)
fix: validate server names in install commands (#3522)
fix: use raw strings for regex in pytest.raises match (#3523)
fix: reject refresh tokens used as Bearer access tokens (#3524)
fix: route ResourcesAsTools/PromptsAsTools through server middleware (#3495)
fix: resolve Pyright "Module is not callable" on @tool, @resource, @prompt decorators (#3540)
fix: filter warnings by message in KEY_PREFIX test (#3549)
fix: suppress output schema for ToolResult subclass annotations (#3548)
fix: increase sleep duration in proxy cache tests (#3567)
fix: store absolute token expiry to prevent stale expires_in on reload (#3572)
fix: preserve tool properties named 'title' during schema compression (#3582)
Fix loopback redirect URI port matching per RFC 8252 §7.3 (#3589)
Fix app tool routing: visibility check and middleware propagation (#3591)
Fix query parameter serialization to respect OpenAPI explode/style settings (#3595)
Fix dev apps form: union types, textarea support, JSON parsing (#3597)
fix(google): replace deprecated /oauth2/v1/tokeninfo with /oauth2/v3/userinfo (#3603)
fix: resolve EntraOBOToken dependency injection through MultiAuth (#3609)
fix(docs): correct misleading stateless_http header (#3622)
fix: filesystem provider import machinery (#3626)
Closes #3625 (issues 2, 3, 6)
fix: recover StdioTransport after subprocess exits (#3630)
fix(server): preserve mounted tool task metadata (#3632)
fix: scope deprecation warning filter to FastMCPDeprecationWarning (#3649)
fix imports, add PrefabAppConfig (#3650)
fix: resolve CurrentFastMCP/ctx.fastmcp to child server in mounted background tasks (#3651)
Fix blocking docs issues: chart imports, Select API, Rx consistency (#3652)
closed by default (#3657)
Fix prompt caching middleware missing wrap/unwrap round-trip (#3666)
fix: serialize object query params per OpenAPI style/explode rules (#3662)
Fixes #2857
fix: HTTP request headers not accessible in background task workers (#3631)
fix: restore HTTP headers in worker execution path for background tasks (#3681)
fix: strip discriminator after dereferencing schemas (#3682)
fix: remove stale ty:ignore directives for ty 0.0.26 (#3684)
Fix docs gaps in app provider pages (#3690)
fix: dev apps log panel UX improvements (#3698)
fix dev server empty string args (#3700)
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

bug Something isn't working. Reports of errors, unexpected behavior, or broken functionality. server Related to FastMCP server implementation or server-side functionality.

Projects

None yet

Development

Successfully merging this pull request may close these issues.

ToolResult subclasses lead to validation errors in fastmcp 3

1 participant