tool-guard-core 0.8.0__tar.gz

tool_guard_core-0.8.0/IMPLEMENTATION_NOTES.md

@@ -0,0 +1,159 @@
+# SDK Implementation Notes
+
+## What was built
+
+Full Python SDK (`toolguard`, initially built as unpublished v0.1.0; PyPI
+distribution `tool-guard-core` begins with the lockstep v0.8.0 release) for
+Tool Guard Core, shipped at `sdk/python/`. It turns the doc-pattern-only
+LangChain/AutoGen/native
+snippets in `docs/integration.md` into shipped, tested, importable code.
+
+## File list
+
+```
+sdk/python/
+├── toolguard/
+│   ├── __init__.py            — public exports
+│   ├── types.py               — dataclasses mirroring Go pkg/domain contracts
+│   ├── client.py              — ToolGuard (CLI + proxy backends)
+│   ├── errors.py              — ToolDenied / ToolEscalated
+│   └── adapters/
+│       ├── __init__.py
+│       ├── langchain.py       — ToolGuardCallbackHandler (BaseCallbackHandler)
+│       ├── autogen.py         — guarded() decorator + register_guarded()
+│       ├── native.py          — guard_tool_calls() (OpenAI + Anthropic)
+│       └── mcp.py             — guard_mcp_tool() decorator
+├── tests/
+│   ├── __init__.py
+│   ├── test_types.py          — JSON round-trip / field-name contract
+│   ├── test_client.py         — CLI mode (exit codes) + proxy mode (httpx mock)
+│   ├── test_langchain.py      — allow/deny/escalate through callback handler
+│   ├── test_autogen.py        — guarded() decorator allow/deny/escalate
+│   ├── test_native.py         — guard_tool_calls() OpenAI + Anthropic shapes
+│   ├── test_mcp.py            — guard_mcp_tool() decorator
+│   ├── test_coverage_extras.py — targeted gap-filling for 90%+ coverage
+│   └── test_contract.py       — real engine contract (builds tg from source)
+├── pyproject.toml             — package metadata; core dep: httpx
+├── README.md                  — per-framework usage examples
+└── IMPLEMENTATION_NOTES.md    — this file
+```
+
+Plus changes in the repo root:
+- `CHANGELOG.md` — entry under `[0.5.0]`
+- `Makefile` — `sdk-test:` target
+- `docs/integration.md` — sections 3.2–3.4 replaced with SDK usage
+
+## Test results
+
+```
+135 passed, 0 failed
+Total coverage: 94.82%  (exceeds the 90% gate)
+```
+
+Post-review update: 2 more passed since the initial 133 — added
+`test_real_dispatcher_propagates_deny` and `test_init_sets_raise_error`
+in `test_langchain.py` (see "Deferred / known gaps" — this closes the
+gap that let the `raise_error` bug ship past the original 133).
+
+Run:
+```bash
+cd sdk/python && .venv/bin/python -m pytest tests/ --cov-fail-under=90 -q
+```
+
+## JSON field names mirrored (for spot-check)
+
+These are verbatim copies of the Go `json:"..."` struct tags from
+`pkg/domain/envelope.go` and `pkg/domain/trace.go`:
+
+**ActionEnvelope**:
+`envelope_id`, `timestamp`, `agent_id`, `session_id`, `org_id`,
+`agent_version`, `framework`, `turn_number`, `department`,
+`tool_name`, `tool_server`, `tool_group`, `parameters`,
+`parameters_redacted`, `context`, `integration_type`,
+`proxy_version`, `tls_verified`
+
+**EnvelopeContext** fields:
+`verified`, `session_state`, `agent_supplied`
+
+**VerifiedContext** fields (selected):
+`customer_tier`, `customer_account_age_days`, `customer_lifetime_value`,
+`customer_order_count`, `order_total`, `order_age_days`, `order_currency`,
+`customer_id`, `order_id`, `order_item_count`, `product_category`,
+`product_category_avg_price`, `return_request_reason`,
+`return_request_status`, `economic_value_impact`,
+`rolling_24h_total`, `rolling_24h_count`, `rolling_7d_total`, `rolling_7d_count`,
+`agent_budget`, `agent_velocity`, `justification`,
+`content_risk`, `content_categories`, `content_classifier_tier`,
+`counter_agent_risk`, `counter_agent_categories`, `counter_agent_reasoning`
+
+**AgentBudgetContext**: `total_limit`, `used_today`, `remaining`, `transactions_today`
+
+**AgentVelocityContext**: `monetary_count_1h`, `monetary_sum_1h`,
+`monetary_count_24h`, `monetary_sum_24h`, `token_count_1h`,
+`token_count_24h`, `llm_cost_usd_1h`, `llm_cost_usd_24h`
+
+**SessionStateContext**: `cumulative_amount`, `actions_in_session`,
+`escalations_in_session`, `denied_in_session`, `tool_sequence`, `amount_trajectory`
+
+**EvaluationResult** (trace.go:165):
+`decision`, `action_taken`, `decision_reason`, `effective_mode`,
+`policies_matched`, `rules_evaluated`, `rules_triggered`,
+`rule_results`, `primary_citation`, `is_near_miss`, `suggested_response`
+
+**Decision** wire values: `"allowed"`, `"denied"`, `"escalated"`, `"flagged"`
+**ActionTaken** wire values: `"allowed"`, `"denied"`, `"escalated"`, `"flagged"`, `"allowed_shadow"`
+
+**RuleResult**: `rule_id`, `rule_name`, `policy_id`, `policy_version`,
+`matched`, `effect`, `severity`, `citation`, `details`
+
+**Citation**: `document_id`, `document_title`, `section`, `page`, `line`, `excerpt`
+
+## CLI backend details
+
+`tg evaluate` reads one policy set (`-policy FILE` or `-policy-dir DIR`) and
+one envelope JSON (`-call`) per invocation. The SDK writes the envelope to a
+temp file and passes its configured policy file or directory to the Go engine
+in one call. Evaluating the complete set together preserves mixed
+shadow/enforcement behavior exactly instead of approximating it in Python.
+
+Exit-code contract (from `cmd/tg/main.go` cmdEvaluate):
+- `0` → allowed / allowed_shadow
+- `3` → denied
+- `4` → escalated
+- `1` → internal error (raises `RuntimeError`)
+- `2` → usage error (raises `RuntimeError`)
+
+## Framework / integration_type stamps
+
+| Adapter | framework | integration_type |
+|---|---|---|
+| Direct `ToolGuard()` | `"sdk"` | `"sdk"` |
+| LangChain | `"langgraph"` | `"langgraph_middleware"` |
+| AutoGen | `"sdk"` | `"sdk"` |
+| Native OpenAI/Anthropic | `"sdk"` | `"sdk"` |
+| MCP | `"mcp"` | `"mcp_proxy"` |
+
+These match the Go constants (`domain.Framework*`, `domain.IntegrationType*`).
+
+## Deferred / known gaps
+
+- **`adapters/langchain.py` line 78–82, 104**: The `on_tool_start` error
+  case when input_str is a non-string type (line 104) and the `except`
+  branch for json.loads failure (lines 77–82) are hard to trigger with
+  simple unit tests.  Covered at 78%; would reach ~85% with integration
+  tests using a real LangChain chain.  Full coverage requires installing
+  langchain-core, which is an optional dep not in the test matrix.
+
+- **`adapters/autogen.py` line 71**: The positional-args path (`_args`
+  key) is an edge case; AutoGen always passes kwargs.  At 93%.
+
+- **`client.py` lines 267–268**: The `httpx.ImportError` message path
+  is not covered because httpx is always installed in the test venv.
+  At 96%.
+
+- **Async adapters**: LangGraph's async tool call path (using
+  `on_tool_start_async`) is not yet implemented.  The synchronous handler
+  covers 95% of real LangChain usage.
+
+- **Go CI integration**: done — `.github/workflows/ci.yml` now has an
+  `sdk-python` job (Python 3.10 + 3.12 matrix) running `make sdk-test`.

tool_guard_core-0.8.0/LICENSE

@@ -0,0 +1,202 @@
+
+                                 Apache License
+                           Version 2.0, January 2004
+                        http://www.apache.org/licenses/
+
+   TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
+
+   1. Definitions.
+
+      "License" shall mean the terms and conditions for use, reproduction,
+      and distribution as defined by Sections 1 through 9 of this document.
+
+      "Licensor" shall mean the copyright owner or entity authorized by
+      the copyright owner that is granting the License.
+
+      "Legal Entity" shall mean the union of the acting entity and all
+      other entities that control, are controlled by, or are under common
+      control with that entity. For the purposes of this definition,
+      "control" means (i) the power, direct or indirect, to cause the
+      direction or management of such entity, whether by contract or
+      otherwise, or (ii) ownership of fifty percent (50%) or more of the
+      outstanding shares, or (iii) beneficial ownership of such entity.
+
+      "You" (or "Your") shall mean an individual or Legal Entity
+      exercising permissions granted by this License.
+
+      "Source" form shall mean the preferred form for making modifications,
+      including but not limited to software source code, documentation
+      source, and configuration files.
+
+      "Object" form shall mean any form resulting from mechanical
+      transformation or translation of a Source form, including but
+      not limited to compiled object code, generated documentation,
+      and conversions to other media types.
+
+      "Work" shall mean the work of authorship, whether in Source or
+      Object form, made available under the License, as indicated by a
+      copyright notice that is included in or attached to the work
+      (an example is provided in the Appendix below).
+
+      "Derivative Works" shall mean any work, whether in Source or Object
+      form, that is based on (or derived from) the Work and for which the
+      editorial revisions, annotations, elaborations, or other modifications
+      represent, as a whole, an original work of authorship. For the purposes
+      of this License, Derivative Works shall not include works that remain
+      separable from, or merely link (or bind by name) to the interfaces of,
+      the Work and Derivative Works thereof.
+
+      "Contribution" shall mean any work of authorship, including
+      the original version of the Work and any modifications or additions
+      to that Work or Derivative Works thereof, that is intentionally
+      submitted to Licensor for inclusion in the Work by the copyright owner
+      or by an individual or Legal Entity authorized to submit on behalf of
+      the copyright owner. For the purposes of this definition, "submitted"
+      means any form of electronic, verbal, or written communication sent
+      to the Licensor or its representatives, including but not limited to
+      communication on electronic mailing lists, source code control systems,
+      and issue tracking systems that are managed by, or on behalf of, the
+      Licensor for the purpose of discussing and improving the Work, but
+      excluding communication that is conspicuously marked or otherwise
+      designated in writing by the copyright owner as "Not a Contribution."
+
+      "Contributor" shall mean Licensor and any individual or Legal Entity
+      on behalf of whom a Contribution has been received by Licensor and
+      subsequently incorporated within the Work.
+
+   2. Grant of Copyright License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      copyright license to reproduce, prepare Derivative Works of,
+      publicly display, publicly perform, sublicense, and distribute the
+      Work and such Derivative Works in Source or Object form.
+
+   3. Grant of Patent License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      (except as stated in this section) patent license to make, have made,
+      use, offer to sell, sell, import, and otherwise transfer the Work,
+      where such license applies only to those patent claims licensable
+      by such Contributor that are necessarily infringed by their
+      Contribution(s) alone or by combination of their Contribution(s)
+      with the Work to which such Contribution(s) was submitted. If You
+      institute patent litigation against any entity (including a
+      cross-claim or counterclaim in a lawsuit) alleging that the Work
+      or a Contribution incorporated within the Work constitutes direct
+      or contributory patent infringement, then any patent licenses
+      granted to You under this License for that Work shall terminate
+      as of the date such litigation is filed.
+
+   4. Redistribution. You may reproduce and distribute copies of the
+      Work or Derivative Works thereof in any medium, with or without
+      modifications, and in Source or Object form, provided that You
+      meet the following conditions:
+
+      (a) You must give any other recipients of the Work or
+          Derivative Works a copy of this License; and
+
+      (b) You must cause any modified files to carry prominent notices
+          stating that You changed the files; and
+
+      (c) You must retain, in the Source form of any Derivative Works
+          that You distribute, all copyright, patent, trademark, and
+          attribution notices from the Source form of the Work,
+          excluding those notices that do not pertain to any part of
+          the Derivative Works; and
+
+      (d) If the Work includes a "NOTICE" text file as part of its
+          distribution, then any Derivative Works that You distribute must
+          include a readable copy of the attribution notices contained
+          within such NOTICE file, excluding those notices that do not
+          pertain to any part of the Derivative Works, in at least one
+          of the following places: within a NOTICE text file distributed
+          as part of the Derivative Works; within the Source form or
+          documentation, if provided along with the Derivative Works; or,
+          within a display generated by the Derivative Works, if and
+          wherever such third-party notices normally appear. The contents
+          of the NOTICE file are for informational purposes only and
+          do not modify the License. You may add Your own attribution
+          notices within Derivative Works that You distribute, alongside
+          or as an addendum to the NOTICE text from the Work, provided
+          that such additional attribution notices cannot be construed
+          as modifying the License.
+
+      You may add Your own copyright statement to Your modifications and
+      may provide additional or different license terms and conditions
+      for use, reproduction, or distribution of Your modifications, or
+      for any such Derivative Works as a whole, provided Your use,
+      reproduction, and distribution of the Work otherwise complies with
+      the conditions stated in this License.
+
+   5. Submission of Contributions. Unless You explicitly state otherwise,
+      any Contribution intentionally submitted for inclusion in the Work
+      by You to the Licensor shall be under the terms and conditions of
+      this License, without any additional terms or conditions.
+      Notwithstanding the above, nothing herein shall supersede or modify
+      the terms of any separate license agreement you may have executed
+      with Licensor regarding such Contributions.
+
+   6. Trademarks. This License does not grant permission to use the trade
+      names, trademarks, service marks, or product names of the Licensor,
+      except as required for reasonable and customary use in describing the
+      origin of the Work and reproducing the content of the NOTICE file.
+
+   7. Disclaimer of Warranty. Unless required by applicable law or
+      agreed to in writing, Licensor provides the Work (and each
+      Contributor provides its Contributions) on an "AS IS" BASIS,
+      WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
+      implied, including, without limitation, any warranties or conditions
+      of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
+      PARTICULAR PURPOSE. You are solely responsible for determining the
+      appropriateness of using or redistributing the Work and assume any
+      risks associated with Your exercise of permissions under this License.
+
+   8. Limitation of Liability. In no event and under no legal theory,
+      whether in tort (including negligence), contract, or otherwise,
+      unless required by applicable law (such as deliberate and grossly
+      negligent acts) or agreed to in writing, shall any Contributor be
+      liable to You for damages, including any direct, indirect, special,
+      incidental, or consequential damages of any character arising as a
+      result of this License or out of the use or inability to use the
+      Work (including but not limited to damages for loss of goodwill,
+      work stoppage, computer failure or malfunction, or any and all
+      other commercial damages or losses), even if such Contributor
+      has been advised of the possibility of such damages.
+
+   9. Accepting Warranty or Additional Liability. While redistributing
+      the Work or Derivative Works thereof, You may choose to offer,
+      and charge a fee for, acceptance of support, warranty, indemnity,
+      or other liability obligations and/or rights consistent with this
+      License. However, in accepting such obligations, You may act only
+      on Your own behalf and on Your sole responsibility, not on behalf
+      of any other Contributor, and only if You agree to indemnify,
+      defend, and hold each Contributor harmless for any liability
+      incurred by, or claims asserted against, such Contributor by reason
+      of your accepting any such warranty or additional liability.
+
+   END OF TERMS AND CONDITIONS
+
+   APPENDIX: How to apply the Apache License to your work.
+
+      To apply the Apache License to your work, attach the following
+      boilerplate notice, with the fields enclosed by brackets "[]"
+      replaced with your own identifying information. (Don't include
+      the brackets!)  The text should be enclosed in the appropriate
+      comment syntax for the file format. We also recommend that a
+      file or class name and description of purpose be included on the
+      same "printed page" as the copyright notice for easier
+      identification within third-party archives.
+
+   Copyright [yyyy] [name of copyright owner]
+
+   Licensed under the Apache License, Version 2.0 (the "License");
+   you may not use this file except in compliance with the License.
+   You may obtain a copy of the License at
+
+       http://www.apache.org/licenses/LICENSE-2.0
+
+   Unless required by applicable law or agreed to in writing, software
+   distributed under the License is distributed on an "AS IS" BASIS,
+   WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+   See the License for the specific language governing permissions and
+   limitations under the License.

tool_guard_core-0.8.0/PKG-INFO

@@ -0,0 +1,318 @@
+Metadata-Version: 2.4
+Name: tool-guard-core
+Version: 0.8.0
+Summary: Python SDK for Tool Guard Core — universal AI agent governance
+Author: DIMAGGI AI
+License-Expression: Apache-2.0
+Project-URL: Homepage, https://dimaggi.ai/tool-guard
+Project-URL: Documentation, https://github.com/dimaggi-ai/tool-guard-core/tree/main/sdk/python
+Project-URL: Source, https://github.com/dimaggi-ai/tool-guard-core
+Project-URL: Issues, https://github.com/dimaggi-ai/tool-guard-core/issues
+Keywords: ai,agents,governance,policy,tool-guard
+Classifier: Development Status :: 4 - Beta
+Classifier: Intended Audience :: Developers
+Classifier: Programming Language :: Python :: 3
+Classifier: Programming Language :: Python :: 3.10
+Classifier: Programming Language :: Python :: 3.11
+Classifier: Programming Language :: Python :: 3.12
+Classifier: Topic :: Software Development :: Libraries :: Python Modules
+Requires-Python: >=3.10
+Description-Content-Type: text/markdown
+License-File: LICENSE
+Requires-Dist: httpx>=0.27
+Provides-Extra: langchain
+Requires-Dist: langchain-core>=0.3; extra == "langchain"
+Provides-Extra: autogen
+Requires-Dist: pyautogen>=0.2; extra == "autogen"
+Provides-Extra: openai
+Requires-Dist: openai>=1.0; extra == "openai"
+Provides-Extra: anthropic
+Requires-Dist: anthropic>=0.30; extra == "anthropic"
+Provides-Extra: dev
+Requires-Dist: pytest>=8; extra == "dev"
+Requires-Dist: pytest-cov>=5; extra == "dev"
+Requires-Dist: pytest-asyncio>=0.24; extra == "dev"
+Requires-Dist: httpx>=0.27; extra == "dev"
+Requires-Dist: langchain-core>=0.3; extra == "dev"
+Dynamic: license-file
+
+# toolguard — Python SDK for Tool Guard Core
+
+Universal AI agent governance: evaluate any tool call against configurable
+policies before it executes.  Drop-in adapters for LangChain, AutoGen,
+native OpenAI/Anthropic tool use, and MCP.  Two backends — CLI (`tg evaluate`)
+and HTTP proxy (`tg-proxy /evaluate`) — so you can run locally or in production.
+
+## Install
+
+The PyPI distribution is named `tool-guard-core` because the unqualified
+`toolguard` project belongs to an unrelated package. The import remains
+`toolguard`. Tagged releases beginning with v0.8.0 install with:
+
+```bash
+pip install "tool-guard-core"                    # core only (httpx)
+pip install "tool-guard-core[langchain]"         # + langchain-core
+pip install "tool-guard-core[autogen]"           # + pyautogen
+pip install "tool-guard-core[openai]"            # + openai SDK
+pip install "tool-guard-core[anthropic]"         # + anthropic SDK
+```
+
+Before v0.8.0, or when testing an unreleased checkout, use
+`pip install "./sdk/python[langchain]"` from the repository root.
+
+## Quick start
+
+```python
+from toolguard import ToolGuard, ToolDenied
+
+guard = ToolGuard(
+    mode="proxy",                        # or "cli" to run tg locally
+    proxy_url="http://localhost:9090",
+    agent_id="my-agent",
+    org_id="acme",
+)
+
+try:
+    result = guard.evaluate("issue_refund", {"amount": 1000})
+    # allowed — call the real tool
+except ToolDenied as exc:
+    # exc.result.decision_reason, exc.result.suggested_response
+    print("Blocked:", exc)
+```
+
+### CLI mode (no running proxy)
+
+```python
+guard = ToolGuard(
+    mode="cli",
+    tg_bin="/usr/local/bin/tg",          # or just "tg" if on PATH
+    policy_dir="/etc/tg/policies",       # directory of *.yaml policies
+    agent_id="my-agent",
+    org_id="acme",
+)
+result = guard.evaluate_raw("issue_refund", {"amount": 100})
+print(result.decision)  # "allowed"
+```
+
+CLI mode evaluates the whole policy directory as one set. Each policy's YAML
+mode owns its contribution: a `mode: shadow` policy remains observe-only under
+the CLI's default `enforcement` call-site mode, while a `mode: enforcement`
+policy cannot be downgraded by a shadow default. Mixed-mode behavior is pinned
+against the real Go engine in `tests/test_contract.py`.
+
+`evaluate()` raises `ToolDenied` or `ToolEscalated` on a block.
+`evaluate_raw()` always returns the `EvaluationResult` without raising.
+
+---
+
+## LangChain / LangGraph
+
+```python
+from toolguard import ToolGuard
+from toolguard.adapters.langchain import ToolGuardCallbackHandler
+from langchain_openai import ChatOpenAI
+from langchain.agents import AgentExecutor
+
+guard = ToolGuard(mode="proxy", proxy_url="http://localhost:9090",
+                  agent_id="lc-agent", org_id="acme")
+
+handler = ToolGuardCallbackHandler(
+    guard,
+    tool_groups={"issue_refund": "monetary_outflow", "search": "read_ops"},
+)
+
+executor = AgentExecutor(agent=agent, tools=tools, callbacks=[handler])
+# Every tool call goes through policy evaluation before execution.
+```
+
+The handler raises `ToolDenied` / `ToolEscalated` on a block, stopping
+the chain.  Flagged calls (recorded near-misses) proceed.
+
+---
+
+## Microsoft AutoGen
+
+```python
+from toolguard import ToolGuard
+from toolguard.adapters.autogen import guarded
+
+guard = ToolGuard(mode="proxy", proxy_url="http://localhost:9090",
+                  agent_id="autogen-agent", org_id="acme")
+
+@guarded(guard, tool_name="issue_refund", tool_group="monetary_outflow")
+def issue_refund(order_id: str, amount: float) -> dict:
+    return {"status": "refunded", "amount": amount}
+
+# Register with AutoGen as normal:
+assistant.register_for_execution(name="issue_refund")(issue_refund)
+```
+
+On deny/escalate, `guarded` returns `{"error": "denied", "reason": "...",
+"suggested_response": "..."}` — the conversation loop surfaces it to the LLM.
+
+---
+
+## Native OpenAI tool use
+
+```python
+import openai
+from toolguard import ToolGuard
+from toolguard.adapters.native import guard_tool_calls
+
+client_oai = openai.OpenAI()
+guard = ToolGuard(mode="proxy", proxy_url="http://localhost:9090",
+                  agent_id="openai-agent", org_id="acme")
+
+response = client_oai.chat.completions.create(
+    model="gpt-4o",
+    messages=messages,
+    tools=tools,
+)
+
+allowed, denied = guard_tool_calls(
+    response.choices[0].message.tool_calls,
+    guard,
+    provider="openai",
+    tool_groups={"issue_refund": "monetary_outflow"},
+)
+
+tool_results = []
+for item in allowed:
+    call = item["call"]
+    output = MY_TOOLS[call.function.name](**json.loads(call.function.arguments))
+    tool_results.append({"role": "tool", "tool_call_id": call.id, "content": str(output)})
+for item in denied:
+    tool_results.append(item["tool_result"])  # role=tool, error message
+```
+
+---
+
+## Native Anthropic tool use
+
+```python
+import anthropic
+from toolguard import ToolGuard
+from toolguard.adapters.native import guard_tool_calls
+
+client_ant = anthropic.Anthropic()
+guard = ToolGuard(mode="proxy", proxy_url="http://localhost:9090",
+                  agent_id="anthropic-agent", org_id="acme")
+
+response = client_ant.messages.create(
+    model="claude-opus-4-5",
+    max_tokens=4096,
+    tools=tools,
+    messages=messages,
+)
+
+tool_use_blocks = [b for b in response.content if b.type == "tool_use"]
+allowed, denied = guard_tool_calls(tool_use_blocks, guard, provider="anthropic")
+
+tool_results = []
+for item in allowed:
+    call = item["call"]
+    output = MY_TOOLS[call.name](**call.input)
+    tool_results.append({"type": "tool_result", "tool_use_id": call.id, "content": str(output)})
+for item in denied:
+    tool_results.append(item["tool_result"])  # type=tool_result, is_error=True
+```
+
+---
+
+## MCP server
+
+```python
+from toolguard import ToolGuard
+from toolguard.adapters.mcp import guard_mcp_tool
+
+guard = ToolGuard(mode="proxy", proxy_url="http://localhost:9090",
+                  agent_id="mcp-server", org_id="acme")
+
+@guard_mcp_tool(guard, tool_group="data_access")
+def call_tool(tool_name: str, parameters: dict, **kwargs) -> dict:
+    return TOOLS[tool_name](**parameters)
+```
+
+---
+
+## Errors
+
+| Exception | When raised |
+|---|---|
+| `ToolDenied` | `action_taken` is anything other than `allowed`, `allowed_shadow`, `flagged`, or `escalated` (i.e. the call was actually blocked) |
+| `ToolEscalated` | `action_taken == "escalated"` (wait for human approval) |
+
+Raised on `action_taken`, not `decision` — deliberately. In shadow mode
+the engine reports `decision="denied"` (what *would* happen) alongside
+`action_taken="allowed_shadow"` (what *actually* happened, since shadow
+mode never enforces); `evaluate()` never raises for that case, since
+shadow mode exists specifically to observe without blocking.
+
+Both carry a `.result` attribute with the full `EvaluationResult`:
+`result.decision_reason`, `result.suggested_response`, `result.rule_results`, etc.
+
+---
+
+## Wire contract
+
+The SDK mirrors the Go `pkg/domain` JSON contract exactly.  Key field names:
+
+**ActionEnvelope** (what the SDK sends to the engine):
+`envelope_id`, `timestamp`, `agent_id`, `session_id`, `org_id`,
+`framework` (`"mcp"` | `"langgraph"` | `"sdk"` | `"unknown"`),
+`tool_name`, `tool_group`, `parameters`, `context.verified`,
+`context.session_state`, `integration_type` (`"mcp_proxy"` |
+`"langgraph_middleware"` | `"sdk"`).
+
+**EvaluationResult** (what the engine returns):
+`decision` (`"allowed"` | `"denied"` | `"escalated"` | `"flagged"`),
+`action_taken`, `decision_reason`, `effective_mode`, `policies_matched`,
+`rules_evaluated`, `rules_triggered`, `rule_results`, `applied_rule_results`,
+`primary_citation`, `applied_primary_citation`, `is_near_miss`,
+`suggested_response`, and the proxy-only optional `decision_receipt`. The
+applied provenance fields explain `action_taken`;
+the aggregate fields can point to stricter shadow-only telemetry.
+
+**DecisionReceipt** (after a successful proxy audit append):
+`receipt_version`, `trace_id`, `trace_hash`, `hash_algorithm`,
+`canonical_trace_version`, `integrity_model`, `decision`, `action_taken`,
+`timestamp`, optional `issuer`, and the opaque `receipt_uri`.
+
+**Escalation** (poll/approve/deny responses): `id`, `state`, timestamps,
+approver fields, `envelope`, `decision`, and an optional
+`resolution_receipt` after the terminal resolution was audited.
+
+## Decision-receipt helper
+
+```python
+from toolguard.adapters.memory import with_receipt_reference
+
+result = guard.evaluate_raw("issue_refund", {"amount": 100})
+event = with_receipt_reference(
+    {"tool": "issue_refund", "outcome": result.action_taken}, result
+)
+```
+
+The helper adds `tool_guard_receipt_uri` only when the proxy supplied a real
+receipt. It performs no storage or network operation and never fabricates a
+fallback identifier. Absence is never authorization. The URN is a
+tamper-evident hash-chain reference, not a signature or fetchable URL; see the
+[full contract](../../docs/integration.md#7-decision-receipts).
+
+---
+
+## Development
+
+```bash
+cd sdk/python
+python3 -m venv .venv
+source .venv/bin/activate
+pip install -e ".[dev]"
+pytest tests/ --cov-fail-under=90
+```
+
+Or via the top-level Makefile:
+
+```bash
+make sdk-test
+```