toolprint 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.
Files changed (54) hide show
  1. toolprint-0.1.0/LICENSE +202 -0
  2. toolprint-0.1.0/PKG-INFO +208 -0
  3. toolprint-0.1.0/README.md +174 -0
  4. toolprint-0.1.0/pyproject.toml +52 -0
  5. toolprint-0.1.0/setup.cfg +4 -0
  6. toolprint-0.1.0/src/toolprint/__init__.py +10 -0
  7. toolprint-0.1.0/src/toolprint/auth.py +219 -0
  8. toolprint-0.1.0/src/toolprint/baseline.py +241 -0
  9. toolprint-0.1.0/src/toolprint/bundle.py +194 -0
  10. toolprint-0.1.0/src/toolprint/canonical.py +278 -0
  11. toolprint-0.1.0/src/toolprint/cli.py +563 -0
  12. toolprint-0.1.0/src/toolprint/connect.py +147 -0
  13. toolprint-0.1.0/src/toolprint/context.py +142 -0
  14. toolprint-0.1.0/src/toolprint/demo.py +182 -0
  15. toolprint-0.1.0/src/toolprint/discover.py +318 -0
  16. toolprint-0.1.0/src/toolprint/drift.py +246 -0
  17. toolprint-0.1.0/src/toolprint/effects.py +215 -0
  18. toolprint-0.1.0/src/toolprint/findings/__init__.py +0 -0
  19. toolprint-0.1.0/src/toolprint/findings/engine.py +376 -0
  20. toolprint-0.1.0/src/toolprint/findings/library.py +161 -0
  21. toolprint-0.1.0/src/toolprint/lexical.py +213 -0
  22. toolprint-0.1.0/src/toolprint/model.py +196 -0
  23. toolprint-0.1.0/src/toolprint/protocol.py +228 -0
  24. toolprint-0.1.0/src/toolprint/py.typed +0 -0
  25. toolprint-0.1.0/src/toolprint/registry.py +188 -0
  26. toolprint-0.1.0/src/toolprint/render/__init__.py +0 -0
  27. toolprint-0.1.0/src/toolprint/render/findings_json.py +147 -0
  28. toolprint-0.1.0/src/toolprint/render/html.py +53 -0
  29. toolprint-0.1.0/src/toolprint/render/jsonout.py +101 -0
  30. toolprint-0.1.0/src/toolprint/render/sarif.py +127 -0
  31. toolprint-0.1.0/src/toolprint/render/text.py +301 -0
  32. toolprint-0.1.0/src/toolprint/tokens.py +98 -0
  33. toolprint-0.1.0/src/toolprint/transport.py +316 -0
  34. toolprint-0.1.0/src/toolprint/vendor/__init__.py +0 -0
  35. toolprint-0.1.0/src/toolprint/vendor/mcp_collect.py +399 -0
  36. toolprint-0.1.0/src/toolprint/viewer/report.html +627 -0
  37. toolprint-0.1.0/src/toolprint.egg-info/PKG-INFO +208 -0
  38. toolprint-0.1.0/src/toolprint.egg-info/SOURCES.txt +52 -0
  39. toolprint-0.1.0/src/toolprint.egg-info/dependency_links.txt +1 -0
  40. toolprint-0.1.0/src/toolprint.egg-info/entry_points.txt +2 -0
  41. toolprint-0.1.0/src/toolprint.egg-info/requires.txt +3 -0
  42. toolprint-0.1.0/src/toolprint.egg-info/top_level.txt +1 -0
  43. toolprint-0.1.0/tests/test_auth.py +119 -0
  44. toolprint-0.1.0/tests/test_bundle.py +294 -0
  45. toolprint-0.1.0/tests/test_canonical.py +174 -0
  46. toolprint-0.1.0/tests/test_context.py +167 -0
  47. toolprint-0.1.0/tests/test_determinism.py +93 -0
  48. toolprint-0.1.0/tests/test_discover.py +107 -0
  49. toolprint-0.1.0/tests/test_drift.py +467 -0
  50. toolprint-0.1.0/tests/test_effects.py +119 -0
  51. toolprint-0.1.0/tests/test_findings.py +238 -0
  52. toolprint-0.1.0/tests/test_html.py +149 -0
  53. toolprint-0.1.0/tests/test_no_secrets.py +136 -0
  54. toolprint-0.1.0/tests/test_protocol.py +155 -0
@@ -0,0 +1,202 @@
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.
176
+
177
+ END OF TERMS AND CONDITIONS
178
+
179
+ APPENDIX: How to apply the Apache License to your work.
180
+
181
+ To apply the Apache License to your work, attach the following
182
+ boilerplate notice, with the fields enclosed by brackets "[]"
183
+ replaced with your own identifying information. (Don't include
184
+ the brackets!) The text should be enclosed in the appropriate
185
+ comment syntax for the file format. We also recommend that a
186
+ file or class name and description of purpose be included on the
187
+ same "printed page" as the copyright notice for easier
188
+ identification within third-party archives.
189
+
190
+ Copyright 2026 Keith Foster
191
+
192
+ Licensed under the Apache License, Version 2.0 (the "License");
193
+ you may not use this file except in compliance with the License.
194
+ You may obtain a copy of the License at
195
+
196
+ http://www.apache.org/licenses/LICENSE-2.0
197
+
198
+ Unless required by applicable law or agreed to in writing, software
199
+ distributed under the License is distributed on an "AS IS" BASIS,
200
+ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
201
+ See the License for the specific language governing permissions and
202
+ limitations under the License.
@@ -0,0 +1,208 @@
1
+ Metadata-Version: 2.4
2
+ Name: toolprint
3
+ Version: 0.1.0
4
+ Summary: What your agent's tool surface costs, and what it can do. Offline MCP assessment: context cost, capability risk, and definition drift.
5
+ Author: Keith Foster
6
+ License-Expression: Apache-2.0
7
+ Project-URL: Homepage, https://github.com/dit4e/toolprint
8
+ Project-URL: Repository, https://github.com/dit4e/toolprint
9
+ Project-URL: Issues, https://github.com/dit4e/toolprint/issues
10
+ Project-URL: Drift dataset, https://github.com/dit4e/toolprint-watch
11
+ Keywords: mcp,model-context-protocol,security,ai-agents,llm,drift-detection,sarif,context-window,supply-chain
12
+ Classifier: Development Status :: 3 - Alpha
13
+ Classifier: Environment :: Console
14
+ Classifier: Intended Audience :: Developers
15
+ Classifier: Intended Audience :: System Administrators
16
+ Classifier: Operating System :: OS Independent
17
+ Classifier: Programming Language :: Python :: 3
18
+ Classifier: Programming Language :: Python :: 3 :: Only
19
+ Classifier: Programming Language :: Python :: 3.9
20
+ Classifier: Programming Language :: Python :: 3.10
21
+ Classifier: Programming Language :: Python :: 3.11
22
+ Classifier: Programming Language :: Python :: 3.12
23
+ Classifier: Programming Language :: Python :: 3.13
24
+ Classifier: Topic :: Security
25
+ Classifier: Topic :: Software Development :: Quality Assurance
26
+ Classifier: Topic :: System :: Monitoring
27
+ Classifier: Typing :: Typed
28
+ Requires-Python: >=3.9
29
+ Description-Content-Type: text/markdown
30
+ License-File: LICENSE
31
+ Provides-Extra: tokens
32
+ Requires-Dist: tiktoken>=0.5; extra == "tokens"
33
+ Dynamic: license-file
34
+
35
+ # toolprint
36
+
37
+ **What your agent's tool surface costs, and what it can do.**
38
+
39
+ Point it at a laptop and get back the number of tokens your MCP tools consume on
40
+ every request before you type anything, how many of them can delete, send or
41
+ spend, and how many of those sit behind no authentication.
42
+
43
+ Offline. No account. Nothing is transmitted.
44
+
45
+ ```bash
46
+ uvx toolprint scan # inventory, no network, no subprocesses
47
+ uvx toolprint scan --connect # retrieve live tool definitions
48
+ uvx toolprint scan --connect --html report.html
49
+ ```
50
+
51
+ ---
52
+
53
+ ## Why the numbers are per-context
54
+
55
+ A conversation loads exactly one client in exactly one project. That pairing is
56
+ the only unit in which "what do my tools cost per conversation" has an answer.
57
+
58
+ Scope precedence is real and mostly invisible: Claude Code resolves local over
59
+ project over user, Cursor and VS Code resolve project over user, and in every
60
+ case the *whole* entry from the winning scope is used — nothing is merged. So a
61
+ machine with 19 configured entries may resolve to 11 distinct servers, of which
62
+ one context loads 8. The other 11 entries are not a conservative overestimate;
63
+ they describe no state the system is ever in. Entries that lose are reported as
64
+ shadowed configuration, because editing one has no effect and produces no
65
+ warning.
66
+
67
+ ## What it reports
68
+
69
+ | | |
70
+ |---|---|
71
+ | **Cost** | Per-tool and per-context token counts, as a share of your window |
72
+ | **Capability** | Every tool classified `read` / `write` / `external` / `irreversible` |
73
+ | **Mislabelled tools** | Tools whose safety annotation their own name or schema contradicts |
74
+ | **Auth posture** | `none`, `env_var`, `literal_secret`, `helper_command`, `oauth` |
75
+ | **Hygiene** | Credentials stored literally in config files, shadowed entries, unreachable servers |
76
+
77
+ Findings carry stable ids (`AUTH-001`, `COST-001`, `HYG-002`, …) with fixed
78
+ titles and standard remediation, so two reports are comparable.
79
+
80
+ ## Output
81
+
82
+ Everything renders from `findings.json`; nothing renders from raw collected data.
83
+
84
+ ```bash
85
+ toolprint scan --connect --findings findings.json --html report.html
86
+ toolprint viewer report.html # empty viewer; drop any findings.json on it
87
+ toolprint demo demo.html # synthetic report, to see the output first
88
+ ```
89
+
90
+ The HTML report is a single self-contained file. All CSS and JS inline, no CDN,
91
+ no fonts, no analytics, and a Content-Security-Policy of `default-src 'none'`
92
+ with `connect-src 'none'` — it has no capability to transmit what it displays.
93
+ It opens from `file://` and prints cleanly to PDF.
94
+
95
+ ## Drift
96
+
97
+ Once a baseline exists, `check` reports what changed without anyone reviewing it.
98
+
99
+ ```bash
100
+ toolprint baseline --connect --by "$USER" --note "reviewed"
101
+ toolprint check --connect --fail-on high
102
+ toolprint approve --tool contact_delete --by "$USER" --note "expected in 4.2"
103
+ ```
104
+
105
+ Ten rules, most severe first: effect-class escalation and revoked safety
106
+ annotations are `critical`; a description that changed while its schema did not
107
+ is `high`, because that is the rug-pull signature — an attacker rewriting a
108
+ tool's instructions has to leave the schema alone or the tool stops working.
109
+ Invisible characters and cross-server references appearing in text are `high`
110
+ too. Breaking schema changes and new tools are `medium`; additive changes and
111
+ removals are `low`.
112
+
113
+ **Baselines refuse to bless a suspicious state.** Trust-on-first-use means a
114
+ baseline approves whatever is in front of it, so `baseline` inspects the current
115
+ surface for invisible characters, homoglyphs and cross-server shadowing first,
116
+ and will not write a clean baseline over any of them without `--force`.
117
+
118
+ **Exceptions require a reason and an expiry.** A suppression that cannot lapse
119
+ becomes permanent by neglect, and then nobody reads the report at all. Expired
120
+ exceptions stop suppressing and say so in the output.
121
+
122
+ Hashes are per component — description, schema, annotations — rather than one
123
+ blob, because only that split shows the description-moved-schema-didn't
124
+ correlation. Canonicalisation normalises structure and never content:
125
+ `transfer` and `transfеr` (Cyrillic е) hash differently on purpose, so a
126
+ homoglyph substitution is caught rather than normalised away.
127
+
128
+ ## CI
129
+
130
+ ```bash
131
+ toolprint scan --format json --fail-on high
132
+ toolprint check --format sarif --out toolprint.sarif
133
+ ```
134
+
135
+ Exit `0` below the threshold, `1` at or above it, `2` on execution error, `3`
136
+ when the baseline is missing or was written by a different heuristics version.
137
+
138
+ SARIF drops into GitHub code scanning, so findings land as annotations on the
139
+ pull request. There is a ready-made action — see
140
+ [docs/github-action.md](docs/github-action.md).
141
+
142
+ ## Assessment bundles
143
+
144
+ `toolprint kit mcp_collect.py` writes a standalone, dependency-free script under
145
+ 400 lines that someone else can run on their own machine to produce a redacted
146
+ bundle. It has exactly one outbound call — reachable only under `--connect`, and
147
+ only to their own MCP servers. Redaction is an allowlist in one visible
148
+ constant. See [docs/collection-kit-README.md](docs/collection-kit-README.md).
149
+
150
+ ## What it catches, and what it does not
151
+
152
+ **Catches:** context cost and its distribution · unauthenticated servers exposing
153
+ destructive capability · mislabelled tools · credentials in config files ·
154
+ shadowed configuration · servers that fail to start.
155
+
156
+ **Does not catch:** a server that is malicious from the first scan — this
157
+ detects change and misconfiguration, not initial badness · a server doing
158
+ exactly what it advertises · runtime behaviour that changes without the tool
159
+ definition changing · whether a description will actually manipulate a given
160
+ model.
161
+
162
+ It composes with scanners that address the first case rather than replacing
163
+ them.
164
+
165
+ ## Install
166
+
167
+ ```bash
168
+ uvx toolprint scan # no install
169
+ pipx install toolprint
170
+ pip install "toolprint[tokens]" # optional tiktoken for exact token counts
171
+ ```
172
+
173
+ Python 3.9+. The core has no required dependencies. Without `tiktoken`, token
174
+ counts use a character ratio calibrated against real tool definitions (4.55
175
+ chars/token, ~4% median error), and the method used is recorded per server.
176
+
177
+ ## Why it reports what it reports
178
+
179
+ [DESIGN.md](DESIGN.md) covers the decisions most likely to surprise you: why
180
+ cost is per context, why annotations give a floor and a ceiling, why
181
+ canonicalisation refuses to normalise Unicode, and why the drift rules are
182
+ ordered as they are.
183
+
184
+ ## Does drift actually happen?
185
+
186
+ [dit4e/toolprint-watch](https://github.com/dit4e/toolprint-watch) checks 27
187
+ public MCP servers daily against an approved baseline and commits what changed,
188
+ so the git history of that baseline is an open record of how often tool
189
+ definitions move. It is a separate repository so a year of observation commits
190
+ does not bury this one.
191
+
192
+ Worth knowing what it cannot see: of 18 well-known public remote MCP endpoints,
193
+ **two** served `tools/list` without credentials. Remote servers are the category
194
+ where a definition change is least visible to the people using them — no package
195
+ update, no lockfile line — and also the category nobody outside the vendor can
196
+ audit. That is an argument for running `toolprint check` in your own CI, where
197
+ you already hold the credentials, rather than waiting for someone else to notice.
198
+
199
+ ## Status
200
+
201
+ Early, but complete through drift. `scan`, `check`, baselines, approvals,
202
+ bundles, the HTML viewer and SARIF are built and dogfooded. Opt-in peer
203
+ benchmarking is next; `findings.json` already carries the `benchmark` field it
204
+ will populate.
205
+
206
+ ## Licence
207
+
208
+ Apache-2.0.
@@ -0,0 +1,174 @@
1
+ # toolprint
2
+
3
+ **What your agent's tool surface costs, and what it can do.**
4
+
5
+ Point it at a laptop and get back the number of tokens your MCP tools consume on
6
+ every request before you type anything, how many of them can delete, send or
7
+ spend, and how many of those sit behind no authentication.
8
+
9
+ Offline. No account. Nothing is transmitted.
10
+
11
+ ```bash
12
+ uvx toolprint scan # inventory, no network, no subprocesses
13
+ uvx toolprint scan --connect # retrieve live tool definitions
14
+ uvx toolprint scan --connect --html report.html
15
+ ```
16
+
17
+ ---
18
+
19
+ ## Why the numbers are per-context
20
+
21
+ A conversation loads exactly one client in exactly one project. That pairing is
22
+ the only unit in which "what do my tools cost per conversation" has an answer.
23
+
24
+ Scope precedence is real and mostly invisible: Claude Code resolves local over
25
+ project over user, Cursor and VS Code resolve project over user, and in every
26
+ case the *whole* entry from the winning scope is used — nothing is merged. So a
27
+ machine with 19 configured entries may resolve to 11 distinct servers, of which
28
+ one context loads 8. The other 11 entries are not a conservative overestimate;
29
+ they describe no state the system is ever in. Entries that lose are reported as
30
+ shadowed configuration, because editing one has no effect and produces no
31
+ warning.
32
+
33
+ ## What it reports
34
+
35
+ | | |
36
+ |---|---|
37
+ | **Cost** | Per-tool and per-context token counts, as a share of your window |
38
+ | **Capability** | Every tool classified `read` / `write` / `external` / `irreversible` |
39
+ | **Mislabelled tools** | Tools whose safety annotation their own name or schema contradicts |
40
+ | **Auth posture** | `none`, `env_var`, `literal_secret`, `helper_command`, `oauth` |
41
+ | **Hygiene** | Credentials stored literally in config files, shadowed entries, unreachable servers |
42
+
43
+ Findings carry stable ids (`AUTH-001`, `COST-001`, `HYG-002`, …) with fixed
44
+ titles and standard remediation, so two reports are comparable.
45
+
46
+ ## Output
47
+
48
+ Everything renders from `findings.json`; nothing renders from raw collected data.
49
+
50
+ ```bash
51
+ toolprint scan --connect --findings findings.json --html report.html
52
+ toolprint viewer report.html # empty viewer; drop any findings.json on it
53
+ toolprint demo demo.html # synthetic report, to see the output first
54
+ ```
55
+
56
+ The HTML report is a single self-contained file. All CSS and JS inline, no CDN,
57
+ no fonts, no analytics, and a Content-Security-Policy of `default-src 'none'`
58
+ with `connect-src 'none'` — it has no capability to transmit what it displays.
59
+ It opens from `file://` and prints cleanly to PDF.
60
+
61
+ ## Drift
62
+
63
+ Once a baseline exists, `check` reports what changed without anyone reviewing it.
64
+
65
+ ```bash
66
+ toolprint baseline --connect --by "$USER" --note "reviewed"
67
+ toolprint check --connect --fail-on high
68
+ toolprint approve --tool contact_delete --by "$USER" --note "expected in 4.2"
69
+ ```
70
+
71
+ Ten rules, most severe first: effect-class escalation and revoked safety
72
+ annotations are `critical`; a description that changed while its schema did not
73
+ is `high`, because that is the rug-pull signature — an attacker rewriting a
74
+ tool's instructions has to leave the schema alone or the tool stops working.
75
+ Invisible characters and cross-server references appearing in text are `high`
76
+ too. Breaking schema changes and new tools are `medium`; additive changes and
77
+ removals are `low`.
78
+
79
+ **Baselines refuse to bless a suspicious state.** Trust-on-first-use means a
80
+ baseline approves whatever is in front of it, so `baseline` inspects the current
81
+ surface for invisible characters, homoglyphs and cross-server shadowing first,
82
+ and will not write a clean baseline over any of them without `--force`.
83
+
84
+ **Exceptions require a reason and an expiry.** A suppression that cannot lapse
85
+ becomes permanent by neglect, and then nobody reads the report at all. Expired
86
+ exceptions stop suppressing and say so in the output.
87
+
88
+ Hashes are per component — description, schema, annotations — rather than one
89
+ blob, because only that split shows the description-moved-schema-didn't
90
+ correlation. Canonicalisation normalises structure and never content:
91
+ `transfer` and `transfеr` (Cyrillic е) hash differently on purpose, so a
92
+ homoglyph substitution is caught rather than normalised away.
93
+
94
+ ## CI
95
+
96
+ ```bash
97
+ toolprint scan --format json --fail-on high
98
+ toolprint check --format sarif --out toolprint.sarif
99
+ ```
100
+
101
+ Exit `0` below the threshold, `1` at or above it, `2` on execution error, `3`
102
+ when the baseline is missing or was written by a different heuristics version.
103
+
104
+ SARIF drops into GitHub code scanning, so findings land as annotations on the
105
+ pull request. There is a ready-made action — see
106
+ [docs/github-action.md](docs/github-action.md).
107
+
108
+ ## Assessment bundles
109
+
110
+ `toolprint kit mcp_collect.py` writes a standalone, dependency-free script under
111
+ 400 lines that someone else can run on their own machine to produce a redacted
112
+ bundle. It has exactly one outbound call — reachable only under `--connect`, and
113
+ only to their own MCP servers. Redaction is an allowlist in one visible
114
+ constant. See [docs/collection-kit-README.md](docs/collection-kit-README.md).
115
+
116
+ ## What it catches, and what it does not
117
+
118
+ **Catches:** context cost and its distribution · unauthenticated servers exposing
119
+ destructive capability · mislabelled tools · credentials in config files ·
120
+ shadowed configuration · servers that fail to start.
121
+
122
+ **Does not catch:** a server that is malicious from the first scan — this
123
+ detects change and misconfiguration, not initial badness · a server doing
124
+ exactly what it advertises · runtime behaviour that changes without the tool
125
+ definition changing · whether a description will actually manipulate a given
126
+ model.
127
+
128
+ It composes with scanners that address the first case rather than replacing
129
+ them.
130
+
131
+ ## Install
132
+
133
+ ```bash
134
+ uvx toolprint scan # no install
135
+ pipx install toolprint
136
+ pip install "toolprint[tokens]" # optional tiktoken for exact token counts
137
+ ```
138
+
139
+ Python 3.9+. The core has no required dependencies. Without `tiktoken`, token
140
+ counts use a character ratio calibrated against real tool definitions (4.55
141
+ chars/token, ~4% median error), and the method used is recorded per server.
142
+
143
+ ## Why it reports what it reports
144
+
145
+ [DESIGN.md](DESIGN.md) covers the decisions most likely to surprise you: why
146
+ cost is per context, why annotations give a floor and a ceiling, why
147
+ canonicalisation refuses to normalise Unicode, and why the drift rules are
148
+ ordered as they are.
149
+
150
+ ## Does drift actually happen?
151
+
152
+ [dit4e/toolprint-watch](https://github.com/dit4e/toolprint-watch) checks 27
153
+ public MCP servers daily against an approved baseline and commits what changed,
154
+ so the git history of that baseline is an open record of how often tool
155
+ definitions move. It is a separate repository so a year of observation commits
156
+ does not bury this one.
157
+
158
+ Worth knowing what it cannot see: of 18 well-known public remote MCP endpoints,
159
+ **two** served `tools/list` without credentials. Remote servers are the category
160
+ where a definition change is least visible to the people using them — no package
161
+ update, no lockfile line — and also the category nobody outside the vendor can
162
+ audit. That is an argument for running `toolprint check` in your own CI, where
163
+ you already hold the credentials, rather than waiting for someone else to notice.
164
+
165
+ ## Status
166
+
167
+ Early, but complete through drift. `scan`, `check`, baselines, approvals,
168
+ bundles, the HTML viewer and SARIF are built and dogfooded. Opt-in peer
169
+ benchmarking is next; `findings.json` already carries the `benchmark` field it
170
+ will populate.
171
+
172
+ ## Licence
173
+
174
+ Apache-2.0.
@@ -0,0 +1,52 @@
1
+ [build-system]
2
+ requires = ["setuptools>=61"]
3
+ build-backend = "setuptools.build_meta"
4
+
5
+ [project]
6
+ name = "toolprint"
7
+ version = "0.1.0"
8
+ description = "What your agent's tool surface costs, and what it can do. Offline MCP assessment: context cost, capability risk, and definition drift."
9
+ readme = "README.md"
10
+ requires-python = ">=3.9"
11
+ license = "Apache-2.0"
12
+ license-files = ["LICENSE"]
13
+ authors = [{ name = "Keith Foster" }]
14
+ keywords = ["mcp", "model-context-protocol", "security", "ai-agents", "llm",
15
+ "drift-detection", "sarif", "context-window", "supply-chain"]
16
+ dependencies = [] # core is stdlib-only; this is a load-bearing claim, not an accident
17
+ classifiers = [
18
+ "Development Status :: 3 - Alpha",
19
+ "Environment :: Console",
20
+ "Intended Audience :: Developers",
21
+ "Intended Audience :: System Administrators",
22
+ "Operating System :: OS Independent",
23
+ "Programming Language :: Python :: 3",
24
+ "Programming Language :: Python :: 3 :: Only",
25
+ "Programming Language :: Python :: 3.9",
26
+ "Programming Language :: Python :: 3.10",
27
+ "Programming Language :: Python :: 3.11",
28
+ "Programming Language :: Python :: 3.12",
29
+ "Programming Language :: Python :: 3.13",
30
+ "Topic :: Security",
31
+ "Topic :: Software Development :: Quality Assurance",
32
+ "Topic :: System :: Monitoring",
33
+ "Typing :: Typed",
34
+ ]
35
+
36
+ [project.urls]
37
+ Homepage = "https://github.com/dit4e/toolprint"
38
+ Repository = "https://github.com/dit4e/toolprint"
39
+ Issues = "https://github.com/dit4e/toolprint/issues"
40
+ "Drift dataset" = "https://github.com/dit4e/toolprint-watch"
41
+
42
+ [project.optional-dependencies]
43
+ tokens = ["tiktoken>=0.5"] # optional accuracy upgrade only; never auto-installed
44
+
45
+ [project.scripts]
46
+ toolprint = "toolprint.cli:main"
47
+
48
+ [tool.setuptools.packages.find]
49
+ where = ["src"]
50
+
51
+ [tool.setuptools.package-data]
52
+ toolprint = ["viewer/*.html", "vendor/*.py", "py.typed"]
@@ -0,0 +1,4 @@
1
+ [egg_info]
2
+ tag_build =
3
+ tag_date = 0
4
+
@@ -0,0 +1,10 @@
1
+ """toolprint - what your agent's tool surface costs, and what it can do.
2
+
3
+ Footprint: the tokens tool definitions consume on every request. Fingerprint:
4
+ what those definitions are, and whether they changed without review.
5
+ """
6
+
7
+ # The public name is spelled here and in cli.py's argparse prog only, which is
8
+ # what made the rename from the mcpdrift placeholder a two-line change.
9
+ TOOL_NAME = "toolprint"
10
+ __version__ = "0.1.0"