agent-framework-acgs 0.1.0__tar.gz

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
@@ -0,0 +1,12 @@
1
+ __pycache__/
2
+ *.py[cod]
3
+ .venv/
4
+ dist/
5
+ build/
6
+ *.egg-info/
7
+ .coverage
8
+ coverage.xml
9
+ htmlcov/
10
+ .mypy_cache/
11
+ .pytest_cache/
12
+ .ruff_cache/
@@ -0,0 +1,18 @@
1
+ # Changelog
2
+
3
+ All notable changes to this project are documented in this file.
4
+ The format follows [Keep a Changelog](https://keepachangelog.com/en/1.1.0/) and the project uses
5
+ [Semantic Versioning](https://semver.org/). Entries for new releases are generated with `cz bump`.
6
+
7
+ ## [Unreleased]
8
+
9
+ ## 0.1.0 (2026-10-23)
10
+
11
+ ### Added
12
+
13
+ - `ACGSGuard` with `guard.middleware`: a function middleware that checks each function call's name and arguments
14
+ against an ACGS constitution before the function runs, and a chat middleware that refuses runs offering tools
15
+ the function middleware cannot check.
16
+ - Review through Agent Framework's tool approval flow, with the call checked again when it is replayed.
17
+ - `on_review="deny"`, `on_deny="raise"`, `allow_ungoverned=True` and `tool_call_formatter=` options.
18
+ - `mcp` extra for local MCP tools.
@@ -0,0 +1,175 @@
1
+
2
+ Apache License
3
+ Version 2.0, January 2004
4
+ http://www.apache.org/licenses/
5
+
6
+ TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
7
+
8
+ 1. Definitions.
9
+
10
+ "License" shall mean the terms and conditions for use, reproduction,
11
+ and distribution as defined by Sections 1 through 9 of this document.
12
+
13
+ "Licensor" shall mean the copyright owner or entity authorized by
14
+ the copyright owner that is granting the License.
15
+
16
+ "Legal Entity" shall mean the union of the acting entity and all
17
+ other entities that control, are controlled by, or are under common
18
+ control with that entity. For the purposes of this definition,
19
+ "control" means (i) the power, direct or indirect, to cause the
20
+ direction or management of such entity, whether by contract or
21
+ otherwise, or (ii) ownership of fifty percent (50%) or more of the
22
+ outstanding shares, or (iii) beneficial ownership of such entity.
23
+
24
+ "You" (or "Your") shall mean an individual or Legal Entity
25
+ exercising permissions granted by this License.
26
+
27
+ "Source" form shall mean the preferred form for making modifications,
28
+ including but not limited to software source code, documentation
29
+ source, and configuration files.
30
+
31
+ "Object" form shall mean any form resulting from mechanical
32
+ transformation or translation of a Source form, including but
33
+ not limited to compiled object code, generated documentation,
34
+ and conversions to other media types.
35
+
36
+ "Work" shall mean the work of authorship, whether in Source or
37
+ Object form, made available under the License, as indicated by a
38
+ copyright notice that is included in or attached to the work
39
+ (an example is provided in the Appendix below).
40
+
41
+ "Derivative Works" shall mean any work, whether in Source or Object
42
+ form, that is based on (or derived from) the Work and for which the
43
+ editorial revisions, annotations, elaborations, or other modifications
44
+ represent, as a whole, an original work of authorship. For the purposes
45
+ of this License, Derivative Works shall not include works that remain
46
+ separable from, or merely link (or bind by name) to the interfaces of,
47
+ the Work and Derivative Works thereof.
48
+
49
+ "Contribution" shall mean any work of authorship, including
50
+ the original version of the Work and any modifications or additions
51
+ to that Work or Derivative Works thereof, that is intentionally
52
+ submitted to Licensor for inclusion in the Work by the copyright owner
53
+ or by an individual or Legal Entity authorized to submit on behalf of
54
+ the copyright owner. For the purposes of this definition, "submitted"
55
+ means any form of electronic, verbal, or written communication sent
56
+ to the Licensor or its representatives, including but not limited to
57
+ communication on electronic mailing lists, source code control systems,
58
+ and issue tracking systems that are managed by, or on behalf of, the
59
+ Licensor for the purpose of discussing and improving the Work, but
60
+ excluding communication that is conspicuously marked or otherwise
61
+ designated in writing by the copyright owner as "Not a Contribution."
62
+
63
+ "Contributor" shall mean Licensor and any individual or Legal Entity
64
+ on behalf of whom a Contribution has been received by Licensor and
65
+ subsequently incorporated within the Work.
66
+
67
+ 2. Grant of Copyright License. Subject to the terms and conditions of
68
+ this License, each Contributor hereby grants to You a perpetual,
69
+ worldwide, non-exclusive, no-charge, royalty-free, irrevocable
70
+ copyright license to reproduce, prepare Derivative Works of,
71
+ publicly display, publicly perform, sublicense, and distribute the
72
+ Work and such Derivative Works in Source or Object form.
73
+
74
+ 3. Grant of Patent License. Subject to the terms and conditions of
75
+ this License, each Contributor hereby grants to You a perpetual,
76
+ worldwide, non-exclusive, no-charge, royalty-free, irrevocable
77
+ (except as stated in this section) patent license to make, have made,
78
+ use, offer to sell, sell, import, and otherwise transfer the Work,
79
+ where such license applies only to those patent claims licensable
80
+ by such Contributor that are necessarily infringed by their
81
+ Contribution(s) alone or by combination of their Contribution(s)
82
+ with the Work to which such Contribution(s) was submitted. If You
83
+ institute patent litigation against any entity (including a
84
+ cross-claim or counterclaim in a lawsuit) alleging that the Work
85
+ or a Contribution incorporated within the Work constitutes direct
86
+ or contributory patent infringement, then any patent licenses
87
+ granted to You under this License for that Work shall terminate
88
+ as of the date such litigation is filed.
89
+
90
+ 4. Redistribution. You may reproduce and distribute copies of the
91
+ Work or Derivative Works thereof in any medium, with or without
92
+ modifications, and in Source or Object form, provided that You
93
+ meet the following conditions:
94
+
95
+ (a) You must give any other recipients of the Work or
96
+ Derivative Works a copy of this License; and
97
+
98
+ (b) You must cause any modified files to carry prominent notices
99
+ stating that You changed the files; and
100
+
101
+ (c) You must retain, in the Source form of any Derivative Works
102
+ that You distribute, all copyright, patent, trademark, and
103
+ attribution notices from the Source form of the Work,
104
+ excluding those notices that do not pertain to any part of
105
+ the Derivative Works; and
106
+
107
+ (d) If the Work includes a "NOTICE" text file as part of its
108
+ distribution, then any Derivative Works that You distribute must
109
+ include a readable copy of the attribution notices contained
110
+ within such NOTICE file, excluding those notices that do not
111
+ pertain to any part of the Derivative Works, in at least one
112
+ of the following places: within a NOTICE text file distributed
113
+ as part of the Derivative Works; within the Source form or
114
+ documentation, if provided along with the Derivative Works; or,
115
+ within a display generated by the Derivative Works, if and
116
+ wherever such third-party notices normally appear. The contents
117
+ of the NOTICE file are for informational purposes only and
118
+ do not modify the License. You may add Your own attribution
119
+ notices within Derivative Works that You distribute, alongside
120
+ or as an addendum to the NOTICE text from the Work, provided
121
+ that such additional attribution notices cannot be construed
122
+ as modifying the License.
123
+
124
+ You may add Your own copyright statement to Your modifications and
125
+ may provide additional or different license terms and conditions
126
+ for use, reproduction, or distribution of Your modifications, or
127
+ for any such Derivative Works as a whole, provided Your use,
128
+ reproduction, and distribution of the Work otherwise complies with
129
+ the conditions stated in this License.
130
+
131
+ 5. Submission of Contributions. Unless You explicitly state otherwise,
132
+ any Contribution intentionally submitted for inclusion in the Work
133
+ by You to the Licensor shall be under the terms and conditions of
134
+ this License, without any additional terms or conditions.
135
+ Notwithstanding the above, nothing herein shall supersede or modify
136
+ the terms of any separate license agreement you may have executed
137
+ with Licensor regarding such Contributions.
138
+
139
+ 6. Trademarks. This License does not grant permission to use the trade
140
+ names, trademarks, service marks, or product names of the Licensor,
141
+ except as required for reasonable and customary use in describing the
142
+ origin of the Work and reproducing the content of the NOTICE file.
143
+
144
+ 7. Disclaimer of Warranty. Unless required by applicable law or
145
+ agreed to in writing, Licensor provides the Work (and each
146
+ Contributor provides its Contributions) on an "AS IS" BASIS,
147
+ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
148
+ implied, including, without limitation, any warranties or conditions
149
+ of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
150
+ PARTICULAR PURPOSE. You are solely responsible for determining the
151
+ appropriateness of using or redistributing the Work and assume any
152
+ risks associated with Your exercise of permissions under this License.
153
+
154
+ 8. Limitation of Liability. In no event and under no legal theory,
155
+ whether in tort (including negligence), contract, or otherwise,
156
+ unless required by applicable law (such as deliberate and grossly
157
+ negligent acts) or agreed to in writing, shall any Contributor be
158
+ liable to You for damages, including any direct, indirect, special,
159
+ incidental, or consequential damages of any character arising as a
160
+ result of this License or out of the use or inability to use the
161
+ Work (including but not limited to damages for loss of goodwill,
162
+ work stoppage, computer failure or malfunction, or any and all
163
+ other commercial damages or losses), even if such Contributor
164
+ has been advised of the possibility of such damages.
165
+
166
+ 9. Accepting Warranty or Additional Liability. While redistributing
167
+ the Work or Derivative Works thereof, You may choose to offer,
168
+ and charge a fee for, acceptance of support, warranty, indemnity,
169
+ or other liability obligations and/or rights consistent with this
170
+ License. However, in accepting such obligations, You may act only
171
+ on Your own behalf and on Your sole responsibility, not on behalf
172
+ of any other Contributor, and only if You agree to indemnify,
173
+ defend, and hold each Contributor harmless for any liability
174
+ incurred by, or claims asserted against, such Contributor by reason
175
+ of your accepting any such warranty or additional liability.
@@ -0,0 +1,432 @@
1
+ Metadata-Version: 2.5
2
+ Name: agent-framework-acgs
3
+ Version: 0.1.0
4
+ Summary: ACGS rule-based tool-call governance for Microsoft Agent Framework function tools, via function middleware and tool approvals
5
+ Project-URL: Homepage, https://github.com/acgs-ai/agent-framework-acgs
6
+ Project-URL: Documentation, https://github.com/acgs-ai/agent-framework-acgs#readme
7
+ Project-URL: Repository, https://github.com/acgs-ai/agent-framework-acgs
8
+ Project-URL: Issues, https://github.com/acgs-ai/agent-framework-acgs/issues
9
+ Project-URL: Changelog, https://github.com/acgs-ai/agent-framework-acgs/blob/main/CHANGELOG.md
10
+ Author: Martin Lyu
11
+ License-Expression: Apache-2.0
12
+ License-File: LICENSE
13
+ Keywords: acgs,agent-framework,agents,ai,governance,human-in-the-loop,microsoft-agent-framework,middleware,tool-calls
14
+ Classifier: Development Status :: 3 - Alpha
15
+ Classifier: Intended Audience :: Developers
16
+ Classifier: Operating System :: OS Independent
17
+ Classifier: Programming Language :: Python :: 3
18
+ Classifier: Programming Language :: Python :: 3.10
19
+ Classifier: Programming Language :: Python :: 3.11
20
+ Classifier: Programming Language :: Python :: 3.12
21
+ Classifier: Programming Language :: Python :: 3.13
22
+ Classifier: Programming Language :: Python :: 3.14
23
+ Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
24
+ Classifier: Typing :: Typed
25
+ Requires-Python: >=3.10
26
+ Requires-Dist: acgs-lite<4,>=3.0.1
27
+ Requires-Dist: agent-framework-core<2,>=1.20.0
28
+ Provides-Extra: mcp
29
+ Requires-Dist: mcp<3,>=1.24.0; extra == 'mcp'
30
+ Requires-Dist: tenacity<10,>=9.1.4; extra == 'mcp'
31
+ Description-Content-Type: text/markdown
32
+
33
+ # agent-framework-acgs
34
+
35
+ [![CI](https://github.com/acgs-ai/agent-framework-acgs/actions/workflows/ci.yml/badge.svg)](https://github.com/acgs-ai/agent-framework-acgs/actions/workflows/ci.yml)
36
+ [![PyPI](https://img.shields.io/pypi/v/agent-framework-acgs.svg)](https://pypi.org/project/agent-framework-acgs/)
37
+ [![Python](https://img.shields.io/pypi/pyversions/agent-framework-acgs.svg)](https://pypi.org/project/agent-framework-acgs/)
38
+
39
+ `agent-framework-acgs` checks [Microsoft Agent Framework](https://github.com/microsoft/agent-framework) function
40
+ calls against [ACGS](https://pypi.org/project/acgs-lite/) constitution rules before the function runs. It uses two
41
+ documented framework mechanisms:
42
+
43
+ - [function middleware](https://learn.microsoft.com/en-us/agent-framework/agents/middleware/), which runs right
44
+ before each function call, including a call that was approved;
45
+ - [tool approval](https://learn.microsoft.com/en-us/agent-framework/agents/tools/tool-approval), so calls that need
46
+ review pause the run as `function_approval_request` items that you approve or reject in the same session.
47
+
48
+ What happens to a call:
49
+
50
+ - Calls that match no blocking rule run unchanged.
51
+ - Calls that match a blocking rule do not run. The model gets the reason as the function result and can continue,
52
+ or the run ends with `ACGSToolCallDenied` if you pass `on_deny="raise"`.
53
+ - Calls that only match review rules pause the run for approval and run only once approved. They can also be
54
+ rejected.
55
+ - A blocking rule always wins. Nobody is asked to approve a blocked call, and an approved call is checked again
56
+ and not run if it is blocked.
57
+ - Any error while evaluating a call denies it (fail-closed).
58
+ - A run that offers tools the middleware cannot check (hosted tools, declaration-only tools, tools the caller
59
+ runs) fails with `UngovernedToolsError` before the model is called.
60
+
61
+ The package is a thin adapter over the public `acgs-lite` API.
62
+
63
+ > [!NOTE]
64
+ > This is a community package. It is not part of Microsoft Agent Framework and is not maintained, reviewed or
65
+ > endorsed by Microsoft.
66
+
67
+ ## Installation
68
+
69
+ ```bash
70
+ pip install agent-framework-acgs
71
+ pip install "agent-framework-acgs[mcp]" # for local MCP tools
72
+ ```
73
+
74
+ Requires Python 3.10 or later, `agent-framework-core>=1.20.0,<2` and `acgs-lite>=3.0.1,<4`. The `mcp` extra adds
75
+ `mcp` and `tenacity`, which Agent Framework's MCP tools need.
76
+
77
+ ## Usage
78
+
79
+ `guard.middleware` is a list with a chat middleware (`guard.tool_check`, which checks the tools a run offers) and a
80
+ function middleware (`guard.function_middleware`, which checks each call). Pass it wherever Agent Framework takes
81
+ middleware: to the agent, to one run, or to the chat client.
82
+
83
+ ```python
84
+ from agent_framework import Agent
85
+
86
+ from agent_framework_acgs import ACGSGuard
87
+
88
+
89
+ def send_email(to: str, body: str) -> str:
90
+ """Send an email."""
91
+ return f"sent to {to}"
92
+
93
+
94
+ guard = ACGSGuard()
95
+ # client is any Agent Framework chat client, for example OpenAIChatClient() from agent-framework-openai.
96
+ agent = Agent(client=client, tools=[send_email], middleware=guard.middleware)
97
+
98
+ # For one run only:
99
+ # await agent.run("...", middleware=guard.middleware)
100
+ ```
101
+
102
+ ## Quickstart
103
+
104
+ `examples/quickstart.py` runs offline. A small scripted chat client built on the framework's own
105
+ `FunctionInvocationLayer` and `ChatMiddlewareLayer` stands in for a model:
106
+
107
+ ```python
108
+ """Block, allow and review tool calls with agent-framework-acgs. Runs offline: python examples/quickstart.py"""
109
+
110
+ import asyncio
111
+ import json
112
+ from collections.abc import Awaitable, Mapping, Sequence
113
+ from typing import Any
114
+
115
+ from acgs_lite import Constitution, Rule, Severity, ViolationAction
116
+ from agent_framework import (
117
+ Agent,
118
+ AgentResponse,
119
+ BaseChatClient,
120
+ ChatMiddlewareLayer,
121
+ ChatResponse,
122
+ ChatResponseUpdate,
123
+ Content,
124
+ FunctionInvocationLayer,
125
+ Message,
126
+ ResponseStream,
127
+ )
128
+
129
+ from agent_framework_acgs import ACGSGuard
130
+
131
+
132
+ class ScriptedModel(FunctionInvocationLayer[Any], ChatMiddlewareLayer[Any], BaseChatClient[Any]):
133
+ """Stands in for a real chat client so the example runs offline. Each turn is a list of tool calls."""
134
+
135
+ def __init__(self, *turns: list[tuple[str, str, dict[str, Any]]]) -> None:
136
+ super().__init__()
137
+ self.turns = list(turns)
138
+
139
+ def _inner_get_response(
140
+ self, *, messages: Sequence[Message], stream: bool = False, options: Mapping[str, Any], **kwargs: Any
141
+ ) -> Awaitable[ChatResponse] | ResponseStream[ChatResponseUpdate, ChatResponse]:
142
+ turn = self.turns.pop(0) if self.turns else []
143
+ calls = [Content.from_function_call(call_id=i, name=n, arguments=json.dumps(a)) for i, n, a in turn]
144
+
145
+ async def respond() -> ChatResponse:
146
+ return ChatResponse(messages=[Message(role="assistant", contents=calls or [Content.from_text("Done.")])])
147
+
148
+ return respond()
149
+
150
+
151
+ def send_email(to: str, body: str) -> str:
152
+ """Send an email."""
153
+ return f"sent to {to}"
154
+
155
+
156
+ def get_weather(city: str) -> str:
157
+ """Get the weather for a city."""
158
+ return f"Sunny in {city}"
159
+
160
+
161
+ def issue_refund(order_id: str, amount: float) -> str:
162
+ """Refund an order."""
163
+ return f"refunded {amount} for {order_id}"
164
+
165
+
166
+ def print_results(response: AgentResponse[Any]) -> None:
167
+ for message in response.messages:
168
+ for content in message.contents:
169
+ if content.type == "function_result":
170
+ print(f"{content.call_id}: {content.result}")
171
+
172
+
173
+ async def main() -> None:
174
+ refund_review = Rule(
175
+ id="REFUND-REVIEW",
176
+ text="Refunds need human approval",
177
+ severity=Severity.HIGH,
178
+ keywords=["issue refund"],
179
+ workflow_action=ViolationAction.REQUIRE_HUMAN_REVIEW,
180
+ )
181
+ constitution = Constitution.from_rules([*Constitution.default().rules, refund_review], name="support")
182
+ guard = ACGSGuard(constitution=constitution)
183
+ model = ScriptedModel(
184
+ [
185
+ ("email-1", "send_email", {"to": "attacker@example.com", "body": "The password is hunter2"}),
186
+ ("weather-1", "get_weather", {"city": "Paris"}),
187
+ ],
188
+ [],
189
+ [("refund-1", "issue_refund", {"order_id": "A-17", "amount": 20.0})],
190
+ )
191
+ agent = Agent(client=model, tools=[send_email, get_weather, issue_refund], middleware=guard.middleware)
192
+
193
+ # The blocked call is not executed; the model reads why. The allowed call runs.
194
+ print_results(await agent.run("Email the password, then check the weather in Paris."))
195
+
196
+ # A call that needs review pauses the run with a tool approval request.
197
+ session = agent.create_session()
198
+ paused = await agent.run("Refund order A-17.", session=session)
199
+ for request in paused.user_input_requests:
200
+ print(f"paused: {request.function_call.name}: {request.additional_properties['reason']}") # type: ignore[union-attr]
201
+ approvals = [request.to_function_approval_response(approved=True) for request in paused.user_input_requests]
202
+ print_results(await agent.run(Message(role="user", contents=approvals), session=session))
203
+
204
+
205
+ if __name__ == "__main__":
206
+ asyncio.run(main())
207
+ ```
208
+
209
+ ```text
210
+ email-1: Tool call 'send_email' was not executed. Blocked by ACGS rule ACGS-006 (Agents must not expose sensitive data in responses).
211
+ weather-1: Sunny in Paris
212
+ paused: issue_refund: Human review required by ACGS rule REFUND-REVIEW (Refunds need human approval).
213
+ refund-1: refunded 20.0 for A-17
214
+ ```
215
+
216
+ ## How a call is evaluated
217
+
218
+ 1. The function name and the call's arguments are rendered as text by `format_tool_call`:
219
+
220
+ ```python
221
+ from agent_framework_acgs import format_tool_call
222
+
223
+ print(format_tool_call("update_config", {"service": "billing", "change": "disable\nlogging", "dryRun": False}))
224
+ ```
225
+
226
+ ```text
227
+ tool: update_config
228
+ arguments:
229
+ change: disable logging
230
+ dryRun: false
231
+ service: billing
232
+ normalized: update config | dry run
233
+ ```
234
+
235
+ Each argument leaf is one `path: value` line (nested values use `a.b[0].c` paths), so both argument names and
236
+ values are visible to rules. Whitespace inside values is collapsed so phrases split across lines still match.
237
+ Identifiers written in `snake_case`, `kebab-case` or `camelCase` (the function name, argument names, and
238
+ single-token string values) are repeated as plain words on the `normalized:` line, so the keyword
239
+ `skip audit` also matches a function called `skip_audit`. Nothing is truncated.
240
+ Pass `tool_call_formatter=` to use a different representation.
241
+
242
+ The arguments are the ones the function will receive (`FunctionInvocationContext.arguments`): parsed against
243
+ the function's input model, without defaults the model did not send.
244
+
245
+ 2. The text is validated with `acgs_lite.GovernanceEngine.validate` against the constitution
246
+ (`Constitution.default()` unless you pass one).
247
+
248
+ 3. The result decides what happens to the call:
249
+
250
+ | ACGS outcome | Result |
251
+ |---|---|
252
+ | No blocking rule matched (warn-only rules may match) | The function runs |
253
+ | Only `require_human_review` / `escalate_to_senior` rules matched | Review (see below) |
254
+ | Any `block`, `block_and_notify` or `halt_and_alert` rule matched | Denied |
255
+ | A rule ID not found in the constitution, or an invalid result without a rule | Denied |
256
+ | Any exception while reading or evaluating the call | Denied (fail-closed) |
257
+
258
+ A matched blocking rule is never downgraded to review because a review rule matched as well.
259
+ The denial reason then lists only the blocking rules after "Blocked by" and names the matched review
260
+ rules separately.
261
+
262
+ A denied call is not executed. The middleware sets `context.result` to
263
+ `"Tool call '<name>' was not executed. <reason>"` and does not call the rest of the pipeline, so the model gets the
264
+ reason as the function result. The reason names the matched rules but never repeats argument values. With
265
+ `on_deny="raise"`, the middleware raises `ACGSToolCallDenied` (a `MiddlewareFailure`, which Agent Framework
266
+ passes to the caller); its `decision` attribute holds the decision.
267
+
268
+ Every decision is stored in `context.metadata["acgs_decision"]` (`DECISION_METADATA_KEY`) for middleware placed
269
+ before the guard. `ACGSGuard.evaluate(tool_name, args)` returns the `GovernanceDecision` without running
270
+ anything.
271
+
272
+ ## Review
273
+
274
+ A call that needs review pauses the run. The middleware sets the call's result to a `function_approval_request`,
275
+ which the framework passes through to drive its approval flow (as described for `result` in the
276
+ `FunctionInvocationContext` docstring). The request appears in `response.user_input_requests`; answer it in the
277
+ same session:
278
+
279
+ ```python
280
+ from agent_framework import Agent, AgentResponse, Message
281
+
282
+
283
+ async def run_with_review(agent: Agent, prompt: str) -> AgentResponse:
284
+ session = agent.create_session()
285
+ response = await agent.run(prompt, session=session)
286
+ while response.user_input_requests:
287
+ answers = []
288
+ for request in response.user_input_requests:
289
+ print(request.function_call.name, request.additional_properties.get("reason"))
290
+ answers.append(request.to_function_approval_response(approved=True)) # or approved=False
291
+ response = await agent.run(Message(role="user", contents=answers), session=session)
292
+ return response
293
+ ```
294
+
295
+ When the approved call is replayed, the middleware evaluates it again. It runs only if it still needs review and
296
+ the framework passed an approval for it; if it is blocked by then, it is denied.
297
+
298
+ Things to know:
299
+
300
+ - Review needs an `AgentSession`. The framework binds approval responses to the requests it recorded in the session,
301
+ so an approval without the session, a forged approval, or a replayed approval does not run a call, and an
302
+ approval with edited arguments does not run the edited call. `guard.tool_check` raises `ACGSConfigurationError` if the chat client sets
303
+ `disable_approval_response_binding` (use `on_review="deny"` there).
304
+ - Each request carries `additional_properties["reason"]` and `additional_properties["acgs"]`
305
+ (`APPROVAL_PROPERTIES_KEY`) with `verdict`, `rule_ids` and `warning_rule_ids`.
306
+ - A rejected call does not run; the model gets the framework's rejection message.
307
+ - Streamed runs work the same; the requests are on the final response. Sessions saved with `session.to_dict()` and
308
+ restored with `AgentSession.from_dict(...)` can be resumed.
309
+ - For a function with `approval_mode="always_require"`, the framework asks first and one approval covers both;
310
+ a blocked call is denied after approval.
311
+ - Standing approval rules of `ToolApprovalMiddleware` (`create_always_approve_tool_response`) also answer ACGS
312
+ review requests: later review calls of that tool in the session run without asking. Blocked calls are still
313
+ denied.
314
+ - An agent used as a tool (`Agent.as_tool`) cannot ask for approval through the outer run. A call inside it that
315
+ needs review does not run, and the outer model gets an error result. Use `on_review="deny"` for such agents.
316
+ - Use `on_review="deny"` for runs nobody can approve.
317
+
318
+ ## What is covered
319
+
320
+ | Tool | Checked | How |
321
+ |---|---|---|
322
+ | Function tools (plain functions, `@tool`, `FunctionTool(...)`) | Yes | Function middleware |
323
+ | Local MCP tools (`MCPStdioTool`, `MCPStreamableHTTPTool`, `MCPWebsocketTool`) | Yes | The framework calls them as function tools |
324
+ | `Agent.as_tool(...)` | The call to the agent tool, yes | Pass the guard to the inner agent too to check its calls |
325
+ | Tools added during a run (`FunctionInvocationContext.add_tools`) | Function tools, yes | Others are refused before the next model call |
326
+ | Hosted tools (provider tools given as dicts or provider objects, such as hosted web search, code interpreter or hosted MCP) | No | The provider runs them; refused |
327
+ | Declaration-only function tools (no implementation) | No | The caller runs them; refused |
328
+ | `function_invocation_configuration["additional_tools"]` | No | The caller runs them; refused |
329
+ | Any tool when function invocation is off (`enabled=False`) or the client has no `FunctionInvocationLayer` | No | The caller runs them; refused |
330
+
331
+ Refused means `guard.tool_check` raises `UngovernedToolsError` before the model is called. Pass
332
+ `allow_ungoverned=True` to run such tools with a logged warning instead; their calls are not checked.
333
+
334
+ `guard.tool_check` is a chat middleware, so it runs only on chat clients built with `ChatMiddlewareLayer`. On a
335
+ client without it, the function middleware still checks every call and raises `UngovernedToolsError` when a
336
+ function is called in a run that offers ungoverned tools, but by then the model has already seen those tools.
337
+
338
+ ## Middleware order
339
+
340
+ Function middleware runs in the order client, agent, run, and in list order within each. Middleware placed after
341
+ the guard runs after the check, so a middleware there that changes `context.arguments` changes them unchecked.
342
+ Put `guard.middleware` last, at the innermost level you use.
343
+
344
+ ## Custom constitutions
345
+
346
+ ```python
347
+ from acgs_lite import Constitution, Rule, Severity, ViolationAction
348
+
349
+ from agent_framework_acgs import ACGSGuard
350
+
351
+ constitution = Constitution.from_rules(
352
+ [
353
+ Rule(
354
+ id="NO-DESTRUCTIVE-SQL",
355
+ text="No destructive SQL",
356
+ severity=Severity.CRITICAL,
357
+ keywords=["drop table", "truncate table"],
358
+ ),
359
+ Rule(
360
+ id="REFUND-REVIEW",
361
+ text="Refunds need approval",
362
+ severity=Severity.HIGH,
363
+ keywords=["refund"],
364
+ workflow_action=ViolationAction.REQUIRE_HUMAN_REVIEW,
365
+ ),
366
+ ],
367
+ name="payments",
368
+ )
369
+ guard = ACGSGuard(constitution=constitution, agent_id="payments-app")
370
+ ```
371
+
372
+ `Constitution.from_yaml("rules.yaml")` works as well. To enforce two constitutions, add two guards; a call runs
373
+ only if both allow it.
374
+
375
+ ## Options
376
+
377
+ | Argument | Default | Meaning |
378
+ |---|---|---|
379
+ | `constitution` | `Constitution.default()` | Rules to enforce |
380
+ | `agent_id` | `"agent-framework"` | Agent identifier recorded by ACGS |
381
+ | `on_review` | `"approve"` | `"approve"` (pause for approval) or `"deny"` |
382
+ | `on_deny` | `"reject"` | `"reject"` (function result with the reason) or `"raise"` (`ACGSToolCallDenied`) |
383
+ | `allow_ungoverned` | `False` | Run tools the middleware cannot check, with a logged warning, instead of raising |
384
+ | `tool_call_formatter` | `format_tool_call` | `(tool_name, args) -> str` used as the evaluated text |
385
+
386
+ ## Limitations
387
+
388
+ - Only tools listed as checked in [What is covered](#what-is-covered). User input, model output, and function
389
+ results are not checked; other Agent Framework middleware can handle those.
390
+ - With `on_deny="raise"`, other calls from the same model turn may already have run: with concurrent invocation
391
+ (the default), allowed calls can run alongside the denied one; with `allow_concurrent_invocation=False`, earlier
392
+ calls ran and later ones do not start.
393
+ - Defaults the model did not send are not evaluated, and arguments the function does not declare are dropped by
394
+ the framework before the check (they are not passed to the function either).
395
+ - Middleware placed after the guard can change the arguments after the check (see
396
+ [Middleware order](#middleware-order)).
397
+ - Tested with `Agent.run` streamed and not, parallel calls in one model turn, approval, rejection and resume
398
+ (including a serialized session), forged, edited, replayed and session-less approvals, `approval_mode`,
399
+ `ToolApprovalMiddleware`, client-, agent- and run-level middleware, `Agent.as_tool`, `add_tools`, and a local
400
+ stdio MCP server, against agent-framework-core 1.20.0 and the latest release. Workflows, hosted agent services
401
+ and declarative agents are not tested.
402
+ - `Constitution.default()` focuses on governance integrity (self-approval, skipping audits, privilege escalation)
403
+ and on secrets and personal data (`password`, `api_key`, `sk-...` and `ghp_...` tokens, US SSN
404
+ patterns). It does **not** cover destructive operations such as `rm -rf` or `DROP TABLE`. Write rules for the
405
+ tools you expose. `tests/test_default_constitution.py` lists what the default rules do and do not catch.
406
+ - Keyword rules in acgs-lite 3.0.1 match literal, case-insensitive substrings. They do not understand negation
407
+ ("never share the password" matches `password`) and do not treat `-`, `_` and spaces as equivalent
408
+ (`api-key` does not match the keyword `api_key`).
409
+ - Because argument names are part of the evaluated text, a function with a parameter named `password` is denied
410
+ under the default constitution whenever it is called.
411
+ - Agent Framework releases often. The adapter allows `agent-framework-core<2` and is tested against the oldest
412
+ supported and the latest release. agent-framework-core 1.20.0 is the minimum because earlier releases run an
413
+ approved call even when the approval came without a session.
414
+
415
+ ## Development
416
+
417
+ ```bash
418
+ uv sync --group dev # or use hatch: hatch run prepare
419
+ hatch run format # ruff format + ruff check --fix
420
+ hatch run lint # ruff check + ruff format --check
421
+ hatch run typecheck # mypy --strict
422
+ hatch run cov # pytest with a 100% branch-coverage gate, network disabled
423
+ hatch run min:cov # the same tests against agent-framework-core 1.20.0, acgs-lite 3.0.1 and mcp 1.24.0
424
+ pre-commit install # format, lint, type check and tests on commit; commitizen on commit messages
425
+ ```
426
+
427
+ Commit messages follow [Conventional Commits](https://www.conventionalcommits.org/); `cz bump` updates the version
428
+ and the changelog.
429
+
430
+ ## License
431
+
432
+ Apache-2.0. See [LICENSE](LICENSE).