agent-core-qa 0.8.2__tar.gz

agent_core_qa-0.8.2/.gitignore

@@ -0,0 +1,245 @@
+# Memory compiler runtime state
+memory-compiler/daily/
+memory-compiler/reports/
+memory-compiler/scripts/state.json
+memory-compiler/scripts/last-flush.json
+memory-compiler/scripts/flush.log
+memory-compiler/scripts/compile.log
+memory-compiler/scripts/session-flush-*.md
+memory-compiler/scripts/flush-context-*.md
+
+# Handoff writer state (placed alongside handoff files)
+**/handoff-state.json
+
+# Git worktrees (superpowers / parallel feature work)
+.worktrees/
+# Native (harness) worktrees
+.claude/worktrees/
+
+# Superpowers brainstorm mockups / scratch
+.superpowers/
+
+# Agent-core runtime state (handoffs, bus.sqlite, transcripts, etc.)
+.agent-core/
+
+# Local scratch / smoke-test artifacts
+.tmp/
+
+# Claude Code per-machine local overrides
+.claude/settings.local.json
+.claude/*.lock
+
+# Byte-compiled / optimized / DLL files
+__pycache__/
+*.py[codz]
+*$py.class
+
+# C extensions
+*.so
+
+# Distribution / packaging
+.Python
+build/
+develop-eggs/
+dist/
+downloads/
+eggs/
+.eggs/
+lib/
+lib64/
+parts/
+sdist/
+var/
+wheels/
+share/python-wheels/
+*.egg-info/
+.installed.cfg
+*.egg
+MANIFEST
+
+# PyInstaller
+#  Usually these files are written by a python script from a template
+#  before PyInstaller builds the exe, so as to inject date/other infos into it.
+*.manifest
+*.spec
+
+# Installer logs
+pip-log.txt
+pip-delete-this-directory.txt
+
+# Unit test / coverage reports
+htmlcov/
+.tox/
+.nox/
+.coverage
+.coverage.*
+.cache
+nosetests.xml
+coverage.xml
+*.cover
+*.py.cover
+.hypothesis/
+.pytest_cache/
+cover/
+
+# Translations
+*.mo
+*.pot
+
+# Django stuff:
+*.log
+local_settings.py
+db.sqlite3
+db.sqlite3-journal
+
+# Flask stuff:
+instance/
+.webassets-cache
+
+# Scrapy stuff:
+.scrapy
+
+# Sphinx documentation
+docs/_build/
+
+# PyBuilder
+.pybuilder/
+target/
+
+# Jupyter Notebook
+.ipynb_checkpoints
+
+# IPython
+profile_default/
+ipython_config.py
+
+# pyenv
+#   For a library or package, you might want to ignore these files since the code is
+#   intended to run in multiple environments; otherwise, check them in:
+# .python-version
+
+# pipenv
+#   According to pypa/pipenv#598, it is recommended to include Pipfile.lock in version control.
+#   However, in case of collaboration, if having platform-specific dependencies or dependencies
+#   having no cross-platform support, pipenv may install dependencies that don't work, or not
+#   install all needed dependencies.
+#Pipfile.lock
+
+# UV
+#   Similar to Pipfile.lock, it is generally recommended to include uv.lock in version control.
+#   This is especially recommended for binary packages to ensure reproducibility, and is more
+#   commonly ignored for libraries.
+#uv.lock
+
+# poetry
+#   Similar to Pipfile.lock, it is generally recommended to include poetry.lock in version control.
+#   This is especially recommended for binary packages to ensure reproducibility, and is more
+#   commonly ignored for libraries.
+#   https://python-poetry.org/docs/basic-usage/#commit-your-poetrylock-file-to-version-control
+#poetry.lock
+#poetry.toml
+
+# pdm
+#   Similar to Pipfile.lock, it is generally recommended to include pdm.lock in version control.
+#   pdm recommends including project-wide configuration in pdm.toml, but excluding .pdm-python.
+#   https://pdm-project.org/en/latest/usage/project/#working-with-version-control
+#pdm.lock
+#pdm.toml
+.pdm-python
+.pdm-build/
+
+# pixi
+#   Similar to Pipfile.lock, it is generally recommended to include pixi.lock in version control.
+#pixi.lock
+#   Pixi creates a virtual environment in the .pixi directory, just like venv module creates one
+#   in the .venv directory. It is recommended not to include this directory in version control.
+.pixi
+
+# PEP 582; used by e.g. github.com/David-OConnor/pyflow and github.com/pdm-project/pdm
+__pypackages__/
+
+# Celery stuff
+celerybeat-schedule
+celerybeat.pid
+
+# SageMath parsed files
+*.sage.py
+
+# Environments
+.env
+.envrc
+.venv
+env/
+venv/
+ENV/
+env.bak/
+venv.bak/
+
+# Spyder project settings
+.spyderproject
+.spyproject
+
+# Rope project settings
+.ropeproject
+
+# mkdocs documentation
+/site
+
+# mypy
+.mypy_cache/
+.dmypy.json
+dmypy.json
+
+# Pyre type checker
+.pyre/
+
+# pytype static type analyzer
+.pytype/
+
+# Cython debug symbols
+cython_debug/
+
+# PyCharm
+#  JetBrains specific template is maintained in a separate JetBrains.gitignore that can
+#  be found at https://github.com/github/gitignore/blob/main/Global/JetBrains.gitignore
+#  and can be added to the global gitignore or merged into this file.  For a more nuclear
+#  option (not recommended) you can uncomment the following to ignore the entire idea folder.
+#.idea/
+
+# Abstra
+# Abstra is an AI-powered process automation framework.
+# Ignore directories containing user credentials, local state, and settings.
+# Learn more at https://abstra.io/docs
+.abstra/
+
+# Visual Studio Code
+#  Visual Studio Code specific template is maintained in a separate VisualStudioCode.gitignore 
+#  that can be found at https://github.com/github/gitignore/blob/main/Global/VisualStudioCode.gitignore
+#  and can be added to the global gitignore or merged into this file. However, if you prefer, 
+#  you could uncomment the following to ignore the entire vscode folder
+# .vscode/
+
+# Ruff stuff:
+.ruff_cache/
+
+# PyPI configuration file
+.pypirc
+
+# Cursor
+#  Cursor is an AI-powered code editor. `.cursorignore` specifies files/directories to
+#  exclude from AI features like autocomplete and code analysis. Recommended for sensitive data
+#  refer to https://docs.cursor.com/context/ignore-files
+.cursorignore
+.cursorindexingignore
+
+# Marimo
+marimo/_static/
+marimo/_lsp/
+__marimo__/
+.gstack/
+
+# macOS Finder metadata (occasionally sneaks in via vendored sources)
+.DS_Store
+
+# Operational docs that must not live in source control
+pepper-vault-plan.txt

agent_core_qa-0.8.2/CHANGELOG.md

@@ -0,0 +1,21 @@
+# Changelog
+
+## [0.8.1](https://github.com/jeffrichley/agent_core/compare/v0.8.0...agent-core-qa-v0.8.1) (2026-07-22)
+
+
+### Bug Fixes
+
+* republish package family to complete rate-limited first publish ([#477](https://github.com/jeffrichley/agent_core/issues/477)) ([32977ac](https://github.com/jeffrichley/agent_core/commit/32977ac78da4a15553329bb4e1fdaa9e894834ad))
+
+## [0.8.0](https://github.com/jeffrichley/agent_core/compare/v0.7.0...agent-core-qa-v0.8.0) (2026-07-21)
+
+
+### Features
+
+* **hatchery:** ungendered templates + .mcp.json generation ([#81](https://github.com/jeffrichley/agent_core/issues/81)/[#82](https://github.com/jeffrichley/agent_core/issues/82)) ([#357](https://github.com/jeffrichley/agent_core/issues/357)) ([f48762d](https://github.com/jeffrichley/agent_core/commit/f48762df78eb12cfa9524e9b5a277741534ea8a7)), closes [#311](https://github.com/jeffrichley/agent_core/issues/311)
+* **release:** configure release-please for 12-package synchronized version train ([#428](https://github.com/jeffrichley/agent_core/issues/428)) ([c1b4097](https://github.com/jeffrichley/agent_core/commit/c1b409765c61d09a6657b0e8b5e43a27cb235b3e))
+
+
+### Bug Fixes
+
+* log or justify bare except-pass swallows + test get_client factory (closes [#408](https://github.com/jeffrichley/agent_core/issues/408)) ([#471](https://github.com/jeffrichley/agent_core/issues/471)) ([e67783c](https://github.com/jeffrichley/agent_core/commit/e67783ccb4dfab6f1783d6a67223f475f81526ea))

agent_core_qa-0.8.2/PKG-INFO

@@ -0,0 +1,9 @@
+Metadata-Version: 2.4
+Name: agent-core-qa
+Version: 0.8.2
+Summary: Release-validation scenario runner for the agent_core daemon (pytest-based)
+Requires-Python: >=3.12
+Requires-Dist: fastmcp>=3.0
+Requires-Dist: httpx>=0.27
+Requires-Dist: pytest-asyncio>=0.23
+Requires-Dist: pytest>=8.0

agent_core_qa-0.8.2/README.md

@@ -0,0 +1,49 @@
+# agent-core-qa
+
+Release-validation scenario runner for the agent_core daemon. **A tool, not a being.**
+
+## What it is
+
+Seven pytest scenarios that exercise a running Phase 3.5 test daemon (`agent-core daemon --instance test`) end-to-end. Each scenario maps to a "v0.3.0 cannot ship if broken" surface: daemon liveness, envelope round-trip, install identity (dynamic keystone), brief framework, scheduler, discord_send, voice synthesis.
+
+## Tool-shaped, not being-shaped
+
+This package is a test runner. It has:
+- No `IDENTITY.md`, `SOUL.md`, `MEMORY.md`, `HEARTBEAT.md`, diary, breadcrumbs, or lessons.
+- No continuity of self — each pytest invocation is fresh.
+- No ethical weight — `rm -rf packages/agent-core-qa/` is functionally identical to deleting any pytest suite.
+- No registration on the bus as a peer; it's an HTTP client that exits when done.
+
+Same shape as `pytest` or `playwright`. Procedural capability through playbooks (scenario functions); nothing for the tool to know about itself.
+
+## Architectural inheritance: standing dynamic surface
+
+From Phase 2.6's PR description, the bug-cadence observation: static artifact analysis caught Bugs 1-2; dynamic real install caught Bug 3. The next failure class — "does the daemon actually start after install" — was named as Phase 2.7 territory.
+
+This package is the standing dynamic surface that catches that failure class (Scenario 1 as autouse fixture precondition) and every subsequent dynamic-only failure class on every release going forward. Phase 3.5 + Phase 2.6 + agent-core-qa is one continuous arc: build the sandbox, fix the install, prove the install still works release over release.
+
+## Runbook
+
+```bash
+# v0.3.0 release-validation runbook
+AGENT_CORE_HOME=~/.agent-core-test agent-core daemon install --instance test --release v0.3.0
+agent-core daemon start --instance test
+cd packages/agent-core-qa
+uv run pytest tests/
+# all 7 pass → safe to refresh prod
+# any fail → fix the regression first
+```
+
+For an already-running test daemon, `agent-core daemon refresh --instance test --release vX.Y.Z` does install + restart in one command.
+
+**Runtime note:** `pytest tests/` triggers Scenario 3 (`test_install_identity_dynamic_keystone`), which performs a SECOND install into a sandbox at `/tmp/qa-<uuid>/` to prove the install path is idempotent against a clean home. On a cold cache the torch download alone can take 5–10 minutes; budget 10–15 minutes for a full validation run on first invocation, 2–3 minutes on subsequent runs once wheels are cached.
+
+## Scenarios
+
+1. `test_daemon_liveness` — precondition (autouse).
+2. `test_envelope_send_receives_ack` — bus round-trip.
+3. `test_install_identity_dynamic_keystone` — install code path identity at runtime (the load-bearing scenario for release validation).
+4. `test_brief_framework_compose_and_submit` — brief framework smoke.
+5. `test_scheduler_create_and_delete_roundtrip` — scheduler round-trip.
+6. `test_discord_send_tool_routes_through_stub` — discord_send tool dispatch with stubbed discord endpoint.
+7. `test_voice_synthesize_returns_audio_bytes` — voice synthesis smoke (no audio-quality validation).

agent_core_qa-0.8.2/VERSION

@@ -0,0 +1 @@
+0.8.1

agent_core_qa-0.8.2/pyproject.toml

@@ -0,0 +1,31 @@
+[project]
+name = "agent-core-qa"
+dynamic = ["version"]
+description = "Release-validation scenario runner for the agent_core daemon (pytest-based)"
+requires-python = ">=3.12"
+dependencies = [
+    "pytest>=8.0",
+    "pytest-asyncio>=0.23",
+    "httpx>=0.27",
+    "fastmcp>=3.0",
+]
+
+[build-system]
+requires = ["hatchling", "uv-dynamic-versioning"]
+build-backend = "hatchling.build"
+
+[tool.hatch.build.targets.wheel]
+packages = ["src/agent_core_qa"]
+
+[tool.hatch.version]
+source = "uv-dynamic-versioning"
+
+[tool.uv-dynamic-versioning]
+fallback-version = "0.0.0"
+
+[tool.uv]
+cache-keys = [{ file = "pyproject.toml" }, { git = { commit = true, tags = true } }]
+
+[tool.pytest.ini_options]
+asyncio_mode = "auto"
+testpaths = ["tests"]

agent_core_qa-0.8.2/src/agent_core_qa/__init__.py

@@ -0,0 +1,4 @@
+"""agent-core-qa — release-validation scenario runner for the agent_core daemon.
+
+Tool-shaped, not being-shaped. See README.md for the constraint set.
+"""

agent_core_qa-0.8.2/src/agent_core_qa/client.py

@@ -0,0 +1,294 @@
+"""DaemonClient — thin HTTP wrapper for the agent_core daemon.
+
+Tool-shaped: stateless across calls, no connection pooling beyond what
+httpx does internally, no retry, no session caching.
+
+Transport discovery (Preflight Step 0, Task 3):
+    The daemon's HTTPHost (Starlette+Uvicorn) mounts only MCPHostable
+    endpoint paths — there is NO /bus/publish or /bus/inbox route.
+    Endpoints that implement MCPHostable (i.e. ClaudeCodeMCPEndpoint)
+    are mounted at /mcp/<name>/ and expose a FastMCP Streamable HTTP
+    server. StubEndpoint is NOT MCPHostable — it has no HTTP surface.
+
+    To publish an envelope from the QA runner, we call the `send` MCP
+    tool on a ClaudeCodeMCPEndpoint (e.g. the `qa` agent endpoint). The
+    tool stamps from_=<agent-name> and dispatches to `stub`. The stub's
+    auto_ack=True calls bus.ack() (marks it handled in persistence) but
+    does NOT publish an Acknowledgment envelope back to the sender.
+
+    Round-trip proof: the `send` tool call returns
+    {"status": "published", "id": <envelope_id>} synchronously after
+    the bus has dispatched and stub has delivered+acked. The bus is
+    single-threaded and dispatches in-process, so delivery is complete
+    before the tool returns. We then call `list_pending` on the same
+    agent endpoint to confirm the queue count=0 (nothing waiting).
+
+    Transport: fastmcp.Client connected to
+    http://<host>:<port>/mcp/<agent>/ (Streamable HTTP). Async methods
+    because fastmcp.Client is an async context manager.
+"""
+
+from __future__ import annotations
+
+import asyncio
+import json
+import socket
+import time
+import urllib.parse
+from collections.abc import Callable
+from typing import Any
+
+try:
+    from fastmcp.client import Client as _FastMCPClient
+
+    _HAS_FASTMCP = True
+except ImportError:  # pragma: no cover
+    _HAS_FASTMCP = False
+
+
+class _FakeTCPResponse:
+    """Mimics a minimal httpx.Response for callers of health_check().
+
+    Used when health_check() falls back to a TCP-connect check because
+    the daemon has no dedicated HTTP liveness route. status_code=200
+    means the port is accepting connections; status_code=503 means it
+    isn't (or the connect timed out).
+    """
+
+    def __init__(self, *, status_code: int, text: str = ""):
+        self.status_code = status_code
+        self.text = text
+
+
+class _MCPToolResult:
+    """Thin wrapper around a fastmcp tool-call result.
+
+    Mimics the httpx.Response API used by callers so scenario code reads
+    the same regardless of whether the underlying transport is raw HTTP
+    or MCP-over-HTTP.  status_code=200 means the tool call succeeded;
+    status_code=500 means the tool raised.
+    """
+
+    def __init__(self, *, status_code: int, data: Any = None, text: str = ""):
+        self.status_code = status_code
+        self._data = data
+        self.text = text
+
+    def json(self) -> Any:
+        return self._data
+
+
+class DaemonClient:
+    """HTTP client for the agent_core daemon.
+
+    Targets the daemon's HTTP API (default http://127.0.0.1:8787 for the
+    Phase 3.5 test instance). One client per scenario; no shared state.
+
+    Transport: MCP tools over FastMCP Streamable HTTP at /mcp/<agent>/.
+    send_envelope() and poll_envelopes() are async — call them with
+    `await` from async test functions (pytest-asyncio asyncio_mode=auto).
+
+    Preflight finding: the daemon's HTTPHost (Starlette+Uvicorn) has no
+    dedicated liveness HTTP route — it only mounts MCP endpoint paths.
+    health_check() therefore uses a TCP-connect check on the bind port
+    rather than a GET to a specific path. A successful TCP connect means
+    the daemon process is up and its HTTP server is accepting connections.
+    """
+
+    def __init__(
+        self,
+        base_url: str,
+        *,
+        timeout: float = 10.0,
+        mcp_agent: str = "qa",
+    ):
+        self.base_url = base_url.rstrip("/")
+        self._timeout = timeout
+        self._mcp_agent = mcp_agent
+
+    @property
+    def _mcp_url(self) -> str:
+        """URL of the agent's MCP endpoint on the daemon."""
+        return f"{self.base_url}/mcp/{self._mcp_agent}/"
+
+    def health_check(self) -> _FakeTCPResponse:
+        """TCP-connect check on the daemon's bind port.
+
+        TCP-connect fallback: the daemon's HTTPHost (Starlette+Uvicorn)
+        exposes no dedicated liveness route — only MCP endpoint mounts.
+        A successful TCP connect on the bind port indicates the HTTP
+        server is accepting connections. Returns a response-like object
+        with status_code=200 on success or status_code=503 on failure.
+        """
+        parsed = urllib.parse.urlparse(self.base_url)
+        host = parsed.hostname or "127.0.0.1"
+        port = parsed.port or 80
+        try:
+            conn = socket.create_connection((host, port), timeout=self._timeout)
+            conn.close()
+            return _FakeTCPResponse(status_code=200, text="TCP connect OK")
+        except OSError as exc:
+            return _FakeTCPResponse(status_code=503, text=str(exc))
+
+    async def send_envelope(self, envelope: dict[str, Any]) -> _MCPToolResult:
+        """Call the `send` MCP tool on the agent's ClaudeCodeMCPEndpoint.
+
+        Transport: fastmcp.Client connected to /mcp/<agent>/ (Streamable
+        HTTP). The `send` tool accepts the envelope fields as kwargs and
+        publishes to the bus. The bus stamps from_=<agent-name>.
+
+        The tool returns {"status": "published", "id": <envelope_id>}
+        synchronously after the bus dispatches and the recipient endpoint
+        delivers+acks. This is the round-trip signal for the stub endpoint
+        (stub auto-acks but does NOT publish an Acknowledgment reply).
+
+        envelope fields consumed:
+            to, kind, payload, correlation_id, in_reply_to, metadata,
+            urgency, expires_at. The id field is advisory (daemon
+            generates a fresh uuid4 hex for each send call).
+        """
+        if not _HAS_FASTMCP:  # pragma: no cover
+            return _MCPToolResult(
+                status_code=500,
+                text="fastmcp not installed; add fastmcp>=3.0 to agent-core-qa deps",
+            )
+
+        try:
+            async with _FastMCPClient(self._mcp_url, timeout=self._timeout) as mcp:
+                result = await mcp.call_tool(
+                    "send",
+                    arguments={
+                        "to": envelope["to"],
+                        "kind": envelope["kind"],
+                        "payload": envelope.get("payload", {}),
+                        "correlation_id": envelope.get("correlation_id"),
+                        "in_reply_to": envelope.get("in_reply_to"),
+                        "metadata": envelope.get("metadata", {}),
+                        "urgency": envelope.get("urgency", "green"),
+                        "expires_at": envelope.get("expires_at"),
+                    },
+                )
+            # fastmcp returns a list of content items; unwrap the first text block.
+            data: Any = None
+            if result and hasattr(result[0], "text"):
+                try:
+                    data = json.loads(result[0].text)
+                except (json.JSONDecodeError, AttributeError):
+                    data = result[0].text if hasattr(result[0], "text") else str(result)
+            return _MCPToolResult(status_code=200, data=data, text=str(data))
+        except Exception as exc:
+            return _MCPToolResult(status_code=500, text=str(exc))
+
+    async def poll_envelopes(
+        self,
+        predicate: Callable[[dict[str, Any]], bool],
+        *,
+        timeout: float = 5.0,
+        interval: float = 0.2,
+    ) -> dict[str, Any] | None:
+        """Poll the agent's pickup queue for an envelope matching predicate.
+
+        Calls the `list_pending` MCP tool on the agent's
+        ClaudeCodeMCPEndpoint in a loop until predicate matches or
+        timeout expires. Returns the first matching envelope dict,
+        or None if nothing matched within timeout.
+
+        Note: for the builtin.stub round-trip, stub does NOT publish an
+        Acknowledgment envelope back to the sender — it only calls
+        bus.ack() (marks the envelope handled in persistence). Therefore
+        list_pending on the sender's queue should return count=0 after a
+        successful send. Use poll_envelopes only when a true reply
+        envelope is expected (e.g. scheduler, discord).
+        """
+        if not _HAS_FASTMCP:  # pragma: no cover
+            return None
+
+        deadline = time.monotonic() + timeout
+        while time.monotonic() < deadline:
+            try:
+                async with _FastMCPClient(self._mcp_url, timeout=self._timeout) as mcp:
+                    result = await mcp.call_tool("list_pending", arguments={})
+
+                data: Any = None
+                if result and hasattr(result[0], "text"):
+                    try:
+                        data = json.loads(result[0].text)
+                    except (json.JSONDecodeError, AttributeError):
+                        # Justified: a malformed tool payload is treated as "no
+                        # data" and the poll loop retries on the next tick.
+                        data = None
+                if isinstance(data, dict):
+                    for item in data.get("items", []):
+                        if predicate(item):
+                            return item
+            except Exception:
+                # Justified: this is a best-effort poll loop over a flaky MCP
+                # connection; transient errors are swallowed so it keeps retrying
+                # until the deadline, then returns None below.
+                pass
+            await asyncio.sleep(interval)
+        return None
+
+    async def list_pending(self) -> dict[str, Any]:
+        """Return the agent's current pending queue snapshot.
+
+        Calls the `list_pending` MCP tool. Returns the raw result dict
+        {"meta": {...}, "items": [...]} or {"meta": {}, "items": []} on
+        failure.
+        """
+        if not _HAS_FASTMCP:  # pragma: no cover
+            return {"meta": {}, "items": []}
+
+        try:
+            async with _FastMCPClient(self._mcp_url, timeout=self._timeout) as mcp:
+                result = await mcp.call_tool("list_pending", arguments={})
+
+            if result and hasattr(result[0], "text"):
+                try:
+                    return json.loads(result[0].text)
+                except (json.JSONDecodeError, AttributeError):
+                    # Justified: a malformed payload falls through to the empty
+                    # snapshot returned below.
+                    pass
+        except Exception:
+            # Justified: list_pending is best-effort; any MCP/transport failure
+            # degrades to the documented empty snapshot returned below.
+            pass
+        return {"meta": {}, "items": []}
+
+    async def call_tool(
+        self,
+        tool_name: str,
+        arguments: dict[str, Any] | None = None,
+    ) -> _MCPToolResult:
+        """Call any MCP tool on the agent's ClaudeCodeMCPEndpoint.
+
+        Generic escape hatch for tools beyond `send` and `list_pending`.
+        Used by Scenario 4 to call `compose_brief` and `submit_brief`
+        on the `qa` endpoint (which has the briefs tools mounted via
+        ``briefs_orchestrator`` wiring in the test daemon's config).
+
+        Returns an _MCPToolResult with status_code=200 on success (the
+        tool returned without raising) and status_code=500 on transport
+        error or tool exception. The parsed result is available via
+        ``.json()``; the raw text via ``.text``.
+        """
+        if not _HAS_FASTMCP:  # pragma: no cover
+            return _MCPToolResult(
+                status_code=500,
+                text="fastmcp not installed; add fastmcp>=3.0 to agent-core-qa deps",
+            )
+
+        try:
+            async with _FastMCPClient(self._mcp_url, timeout=self._timeout) as mcp:
+                result = await mcp.call_tool(tool_name, arguments=arguments or {})
+
+            data: Any = None
+            if result and hasattr(result[0], "text"):
+                try:
+                    data = json.loads(result[0].text)
+                except (json.JSONDecodeError, AttributeError):
+                    data = result[0].text if hasattr(result[0], "text") else str(result)
+            return _MCPToolResult(status_code=200, data=data, text=str(data))
+        except Exception as exc:
+            return _MCPToolResult(status_code=500, text=str(exc))