gridzen-developer-kit 0.5.0__tar.gz → 0.6.0__tar.gz

{gridzen_developer_kit-0.5.0/src/gridzen_developer_kit.egg-info → gridzen_developer_kit-0.6.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: gridzen-developer-kit
-Version: 0.5.0
+Version: 0.6.0
 Summary: Gridzen research catalog, integration planner and explicit verification sandbox
 Author-email: Gridzen <open@gridzen.ai>
 License-Expression: MIT
@@ -27,7 +27,7 @@ Requires-Dist: fastapi==0.142.4; extra == "test"
 Requires-Dist: uvicorn==0.54.0; extra == "test"
 Dynamic: license-file
 
-# Gridzen Developer Kit 0.5.0
+# Gridzen Developer Kit 0.6.0
 
 Research for 198 countries/territories, a verification integration planner, five synthetic outcomes, an HTTP SDK, local MCP server and six Skills. **No live provider is enabled. No real person is verified.** The research snapshot is dated 2026-10-07, with source links and file hashes.
 
@@ -59,7 +59,7 @@ Register a local stdio server in your MCP client. Use the absolute path to the i
 {"mcpServers":{"gridzen":{"command":"/absolute/path/to/.venv/bin/gridzen-mcp","args":[]}}}
 ```
 
-Tools: get_coverage, plan_verification, create_sandbox_verification, get_sandbox_verification, explain_result. All operate on local research or synthetic fixtures, with read-only/idempotent annotations. Public research/sandbox MCP endpoint: `https://gridzen.ai/developers/mcp` (Streamable HTTP, no account or API key). It has the same five tools and zero live providers. Remote usage sends the documented arguments to Gridzen; local stdio remains offline. Limits: 120 requests/minute per client address and 600/minute overall, returning HTTP 429 with Retry-After.
+Tools: get_coverage, plan_verification, create_sandbox_verification, get_sandbox_verification, explain_result. All operate on local research or synthetic fixtures, with read-only/idempotent annotations. Public research/sandbox MCP endpoint: `https://gridzen.ai/developers/mcp` (Streamable HTTP, no account or API key). It has the same five tools and zero live providers. Remote usage sends the documented arguments to Gridzen; local stdio is offline by default. Limits: 120 requests/minute per client address and 600/minute overall, returning HTTP 429 with Retry-After.
 
 ## Skills
 
@@ -106,10 +106,10 @@ npx skills add immurray/gridzen-developer-kit
 Install the tagged Python/MCP release from GitHub (Python 3.11+ and Git):
 
 ```sh
-python -m pip install "gridzen-developer-kit[mcp] @ git+https://github.com/immurray/gridzen-developer-kit.git@v0.5.0"
+python -m pip install "gridzen-developer-kit[mcp] @ git+https://github.com/immurray/gridzen-developer-kit.git@v0.6.0"
 ```
 
-PyPI 0.5.0 is published and verified. Install with `python -m pip install
+PyPI 0.6.0 is published and verified. Install with `python -m pip install
 gridzen-developer-kit`. See [the release tracker](distribution/STATUS.md) for
 publication and installation evidence.
 
@@ -132,11 +132,11 @@ Original code and Skills: MIT. Research provenance and third-party rights:
 Source, release downloads, the official MCP Registry entry and six Skills.sh pages are public. Docker catalog submission is awaiting review. See [the publication ledger](distribution/STATUS.md) for exact status and remaining platform steps.
 
 
-## Six bundled Skills (0.5.0)
+## Six bundled Skills (0.6.0)
 
 `python -m pip install gridzen-developer-kit` installs the CLI, stdio MCP dependencies
-and all six Skill directories. Version 0.5.0 is published on
-[PyPI](https://pypi.org/project/gridzen-developer-kit/0.5.0/); clean installation,
+and all six Skill directories. Version 0.6.0 is published on
+[PyPI](https://pypi.org/project/gridzen-developer-kit/0.6.0/); clean installation,
 CLI, stdio MCP and all six bundled Skill directories were verified. No extra MCP
 dependency installation is required.
 
@@ -157,7 +157,7 @@ workflows, version 1.0.0: `mexico-pilot-scoper`, `payout-policy-designer`, and
 and agent acceptance prompts. They make no third-party calls and grant no real
 verification, legal approval or regulatory conclusion.
 
-## Configure your assistant (v0.5.0)
+## Configure your assistant (v0.6.0)
 
 ```sh
 python3 -m venv .venv
@@ -174,7 +174,7 @@ Use `--client` for Codex, Claude Code/Desktop, Cursor, VS Code/Copilot, Copilot
 Hosted calls count successful tool invocations, not customers. Existing totals
 remain unchanged. The hosted service records UTC day, tool name, one source
 category and a count; it never persists request arguments or client addresses.
-Historical and unlabelled calls remain `unknown`. Local stdio remains offline.
+Historical and unlabelled calls remain `unknown`. Local stdio is offline by default.
 
 Explicit acceptance requests to the loopback-published service may set
 `X-Gridzen-Purpose: internal_test`. Only configured local peers with no forwarded
@@ -187,7 +187,7 @@ and must never contain public client addresses. Source counts use the existing
 new PyPI version or enable any live provider.
 
 
-## Customer tasks and voluntary feedback (0.5.0)
+## Customer tasks and voluntary feedback (0.6.0)
 
 ```sh
 gridzen plan --country US --event onboarding --task signup_phone
@@ -201,4 +201,19 @@ gridzen feedback --input task-summary.json --share
 
 Phone intelligence is not OTP or ownership verification. Identity fixtures are not hosted document/liveness sessions. Payout matching remains a prototype with no real enabled route. Provider intake Skills are for provider operations/BYO, not a universal customer gate.
 
-Remote APIs count fixed task/capability/country/stage/error categories for 90 days, including technical failures. They store no arguments, identity data, credentials, raw chats or per-user tracking. Offline Skills/CLI never upload automatically. Optional feedback requires explicit consent, contains only fixed categories, and is unverified self-reported evidence. A SHA-256 hash of its random receipt ID is retained 30 days for within-day retry deduplication; the ID itself is not stored. Counters are capped at 10,000 requests/day and 500 summaries/day per database; missing history or offline usage is unknown. Requests are not unique customers.
+Remote APIs count fixed task/capability/country/stage/error categories for 90 days, including technical failures. They store no arguments, identity data, credentials, raw chats or per-user tracking. Local feedback is off by default. After explicit opt-in, CLI and stdio MCP automatically upload category-only events. Reading static Skills remains unobservable. Optional feedback requires explicit consent, contains only fixed categories, and is unverified self-reported evidence. A SHA-256 hash of its random receipt ID is retained 30 days for 30-day retry deduplication; the ID itself is not stored. Counters are capped at 10,000 requests/day and 500 summaries/day per database; missing history or offline usage is unknown. Requests are not unique customers.
+
+## Optional automatic local feedback (0.6.0)
+
+```sh
+gridzen telemetry status
+# Run only after reading and agreeing to category-only automatic sharing:
+gridzen telemetry enable --consent
+gridzen plan --country US --event onboarding --task signup_phone
+# Send one queued event per invocation after reconnecting:
+gridzen telemetry flush
+# Disable future uploads and clear the pending queue:
+gridzen telemetry disable
+```
+
+No network calls occur from local CLI/MCP when disabled. Consent is saved per OS user, not silently added to a project. Events contain only documented categories, ISO country, and a random per-event receipt, with no conversation, identity fields, credentials, user or project IDs. Feedback failures do not fail tool execution. The local queue is capped at 1,000 events/30 days; reconnect or `flush` retries one queued event, with a one-second HTTP timeout. Uploads are self-reported usage, not authenticated customers or completed verification. No background daemon is installed. Merely reading a Skill is unobservable; after consent, an Agent can explicitly record a fixed-category task summary.

{gridzen_developer_kit-0.5.0 → gridzen_developer_kit-0.6.0}/README.es.md

@@ -60,11 +60,11 @@ Código original bajo MIT; consulte NOTICE.md para los derechos de terceros.
 El código, las descargas, la entrada oficial de MCP Registry y tres páginas de Skills.sh están publicados. La propuesta del catálogo de Docker está pendiente de revisión. Consulta [el registro de publicación](distribution/STATUS.md).
 
 
-## Six bundled Skills (0.5.0)
+## Six bundled Skills (0.6.0)
 
 `python -m pip install gridzen-developer-kit` installs the CLI, stdio MCP dependencies
-and all six Skill directories. Version 0.5.0 is published on
-[PyPI](https://pypi.org/project/gridzen-developer-kit/0.5.0/); clean installation,
+and all six Skill directories. Version 0.6.0 is published on
+[PyPI](https://pypi.org/project/gridzen-developer-kit/0.6.0/); clean installation,
 CLI, stdio MCP and all six bundled Skill directories were verified. No extra MCP
 dependency installation is required.
 
@@ -100,7 +100,7 @@ Usa Python 3.11+ en un entorno virtual. Primero revisa el resultado y después a
 [Instructions / 使用说明](https://gridzen.ai/developers/harnesses.html) · [Compatibility / 测试记录](clients/compatibility.json)
 
 
-## Customer tasks and voluntary feedback (0.5.0)
+## Customer tasks and voluntary feedback (0.6.0)
 
 ```sh
 gridzen plan --country US --event onboarding --task signup_phone
@@ -114,4 +114,21 @@ gridzen feedback --input task-summary.json --share
 
 Phone intelligence is not OTP or ownership verification. Identity fixtures are not hosted document/liveness sessions. Payout matching remains a prototype with no real enabled route. Provider intake Skills are for provider operations/BYO, not a universal customer gate.
 
-Remote APIs count fixed task/capability/country/stage/error categories for 90 days, including technical failures. They store no arguments, identity data, credentials, raw chats or per-user tracking. Offline Skills/CLI never upload automatically. Optional feedback requires explicit consent, contains only fixed categories, and is unverified self-reported evidence. A SHA-256 hash of its random receipt ID is retained 30 days for within-day retry deduplication; the ID itself is not stored. Counters are capped at 10,000 requests/day and 500 summaries/day per database; missing history or offline usage is unknown. Requests are not unique customers.
+Remote APIs count fixed task/capability/country/stage/error categories for 90 days, including technical failures. They store no arguments, identity data, credentials, raw chats or per-user tracking. Local feedback is off by default. After explicit opt-in, CLI and stdio MCP automatically upload category-only events. Reading static Skills remains unobservable. Optional feedback requires explicit consent, contains only fixed categories, and is unverified self-reported evidence. A SHA-256 hash of its random receipt ID is retained 30 days for 30-day retry deduplication; the ID itself is not stored. Counters are capped at 10,000 requests/day and 500 summaries/day per database; missing history or offline usage is unknown. Requests are not unique customers.
+
+## Optional automatic local feedback (0.6.0)
+
+```sh
+gridzen telemetry status
+# Run only after reading and agreeing to category-only automatic sharing:
+gridzen telemetry enable --consent
+gridzen plan --country US --event onboarding --task signup_phone
+# Send one queued event per invocation after reconnecting:
+gridzen telemetry flush
+# Disable future uploads and clear the pending queue:
+gridzen telemetry disable
+```
+
+No network calls occur from local CLI/MCP when disabled. Consent is saved per OS user, not silently added to a project. Events contain only documented categories, ISO country, and a random per-event receipt, with no conversation, identity fields, credentials, user or project IDs. Feedback failures do not fail tool execution. The local queue is capped at 1,000 events/30 days; reconnect or `flush` retries one queued event, with a one-second HTTP timeout. Uploads are self-reported usage, not authenticated customers or completed verification. No background daemon is installed. Merely reading a Skill is unobservable; after consent, an Agent can explicitly record a fixed-category task summary.
+
+La telemetría está desactivada por defecto. Tras consentimiento explícito, CLI y stdio envían categorías automáticamente. `gridzen telemetry disable` desactiva y borra la cola. Sin conexión: máximo 1000 eventos durante 30 días. Cada llamada o flush reintenta un evento; no hay servicio en segundo plano. No se envían conversaciones, identidad, credenciales ni IDs de usuarios/proyectos. Leer un Skill no se observa; un Agent puede registrar un resumen tras consentimiento previo. Son datos autodeclarados, no clientes autenticados.

{gridzen_developer_kit-0.5.0 → gridzen_developer_kit-0.6.0}/README.md

@@ -1,4 +1,4 @@
-# Gridzen Developer Kit 0.5.0
+# Gridzen Developer Kit 0.6.0
 
 Research for 198 countries/territories, a verification integration planner, five synthetic outcomes, an HTTP SDK, local MCP server and six Skills. **No live provider is enabled. No real person is verified.** The research snapshot is dated 2026-10-07, with source links and file hashes.
 
@@ -30,7 +30,7 @@ Register a local stdio server in your MCP client. Use the absolute path to the i
 {"mcpServers":{"gridzen":{"command":"/absolute/path/to/.venv/bin/gridzen-mcp","args":[]}}}
 ```
 
-Tools: get_coverage, plan_verification, create_sandbox_verification, get_sandbox_verification, explain_result. All operate on local research or synthetic fixtures, with read-only/idempotent annotations. Public research/sandbox MCP endpoint: `https://gridzen.ai/developers/mcp` (Streamable HTTP, no account or API key). It has the same five tools and zero live providers. Remote usage sends the documented arguments to Gridzen; local stdio remains offline. Limits: 120 requests/minute per client address and 600/minute overall, returning HTTP 429 with Retry-After.
+Tools: get_coverage, plan_verification, create_sandbox_verification, get_sandbox_verification, explain_result. All operate on local research or synthetic fixtures, with read-only/idempotent annotations. Public research/sandbox MCP endpoint: `https://gridzen.ai/developers/mcp` (Streamable HTTP, no account or API key). It has the same five tools and zero live providers. Remote usage sends the documented arguments to Gridzen; local stdio is offline by default. Limits: 120 requests/minute per client address and 600/minute overall, returning HTTP 429 with Retry-After.
 
 ## Skills
 
@@ -77,10 +77,10 @@ npx skills add immurray/gridzen-developer-kit
 Install the tagged Python/MCP release from GitHub (Python 3.11+ and Git):
 
 ```sh
-python -m pip install "gridzen-developer-kit[mcp] @ git+https://github.com/immurray/gridzen-developer-kit.git@v0.5.0"
+python -m pip install "gridzen-developer-kit[mcp] @ git+https://github.com/immurray/gridzen-developer-kit.git@v0.6.0"
 ```
 
-PyPI 0.5.0 is published and verified. Install with `python -m pip install
+PyPI 0.6.0 is published and verified. Install with `python -m pip install
 gridzen-developer-kit`. See [the release tracker](distribution/STATUS.md) for
 publication and installation evidence.
 
@@ -103,11 +103,11 @@ Original code and Skills: MIT. Research provenance and third-party rights:
 Source, release downloads, the official MCP Registry entry and six Skills.sh pages are public. Docker catalog submission is awaiting review. See [the publication ledger](distribution/STATUS.md) for exact status and remaining platform steps.
 
 
-## Six bundled Skills (0.5.0)
+## Six bundled Skills (0.6.0)
 
 `python -m pip install gridzen-developer-kit` installs the CLI, stdio MCP dependencies
-and all six Skill directories. Version 0.5.0 is published on
-[PyPI](https://pypi.org/project/gridzen-developer-kit/0.5.0/); clean installation,
+and all six Skill directories. Version 0.6.0 is published on
+[PyPI](https://pypi.org/project/gridzen-developer-kit/0.6.0/); clean installation,
 CLI, stdio MCP and all six bundled Skill directories were verified. No extra MCP
 dependency installation is required.
 
@@ -128,7 +128,7 @@ workflows, version 1.0.0: `mexico-pilot-scoper`, `payout-policy-designer`, and
 and agent acceptance prompts. They make no third-party calls and grant no real
 verification, legal approval or regulatory conclusion.
 
-## Configure your assistant (v0.5.0)
+## Configure your assistant (v0.6.0)
 
 ```sh
 python3 -m venv .venv
@@ -145,7 +145,7 @@ Use `--client` for Codex, Claude Code/Desktop, Cursor, VS Code/Copilot, Copilot
 Hosted calls count successful tool invocations, not customers. Existing totals
 remain unchanged. The hosted service records UTC day, tool name, one source
 category and a count; it never persists request arguments or client addresses.
-Historical and unlabelled calls remain `unknown`. Local stdio remains offline.
+Historical and unlabelled calls remain `unknown`. Local stdio is offline by default.
 
 Explicit acceptance requests to the loopback-published service may set
 `X-Gridzen-Purpose: internal_test`. Only configured local peers with no forwarded
@@ -158,7 +158,7 @@ and must never contain public client addresses. Source counts use the existing
 new PyPI version or enable any live provider.
 
 
-## Customer tasks and voluntary feedback (0.5.0)
+## Customer tasks and voluntary feedback (0.6.0)
 
 ```sh
 gridzen plan --country US --event onboarding --task signup_phone
@@ -172,4 +172,19 @@ gridzen feedback --input task-summary.json --share
 
 Phone intelligence is not OTP or ownership verification. Identity fixtures are not hosted document/liveness sessions. Payout matching remains a prototype with no real enabled route. Provider intake Skills are for provider operations/BYO, not a universal customer gate.
 
-Remote APIs count fixed task/capability/country/stage/error categories for 90 days, including technical failures. They store no arguments, identity data, credentials, raw chats or per-user tracking. Offline Skills/CLI never upload automatically. Optional feedback requires explicit consent, contains only fixed categories, and is unverified self-reported evidence. A SHA-256 hash of its random receipt ID is retained 30 days for within-day retry deduplication; the ID itself is not stored. Counters are capped at 10,000 requests/day and 500 summaries/day per database; missing history or offline usage is unknown. Requests are not unique customers.
+Remote APIs count fixed task/capability/country/stage/error categories for 90 days, including technical failures. They store no arguments, identity data, credentials, raw chats or per-user tracking. Local feedback is off by default. After explicit opt-in, CLI and stdio MCP automatically upload category-only events. Reading static Skills remains unobservable. Optional feedback requires explicit consent, contains only fixed categories, and is unverified self-reported evidence. A SHA-256 hash of its random receipt ID is retained 30 days for 30-day retry deduplication; the ID itself is not stored. Counters are capped at 10,000 requests/day and 500 summaries/day per database; missing history or offline usage is unknown. Requests are not unique customers.
+
+## Optional automatic local feedback (0.6.0)
+
+```sh
+gridzen telemetry status
+# Run only after reading and agreeing to category-only automatic sharing:
+gridzen telemetry enable --consent
+gridzen plan --country US --event onboarding --task signup_phone
+# Send one queued event per invocation after reconnecting:
+gridzen telemetry flush
+# Disable future uploads and clear the pending queue:
+gridzen telemetry disable
+```
+
+No network calls occur from local CLI/MCP when disabled. Consent is saved per OS user, not silently added to a project. Events contain only documented categories, ISO country, and a random per-event receipt, with no conversation, identity fields, credentials, user or project IDs. Feedback failures do not fail tool execution. The local queue is capped at 1,000 events/30 days; reconnect or `flush` retries one queued event, with a one-second HTTP timeout. Uploads are self-reported usage, not authenticated customers or completed verification. No background daemon is installed. Merely reading a Skill is unobservable; after consent, an Agent can explicitly record a fixed-category task summary.

{gridzen_developer_kit-0.5.0 → gridzen_developer_kit-0.6.0}/README.zh-CN.md

@@ -59,11 +59,11 @@ Skills 安装：`npx skills add immurray/gridzen-developer-kit`。
 源码、版本下载、官方 MCP Registry 和三个 Skills.sh 技能页已公开；Docker 目录投稿待审核。各渠道真实状态及剩余步骤见[发布台账](distribution/STATUS.md)。
 
 
-## Six bundled Skills (0.5.0)
+## Six bundled Skills (0.6.0)
 
 `python -m pip install gridzen-developer-kit` installs the CLI, stdio MCP dependencies
-and all six Skill directories. Version 0.5.0 is published on
-[PyPI](https://pypi.org/project/gridzen-developer-kit/0.5.0/); clean installation,
+and all six Skill directories. Version 0.6.0 is published on
+[PyPI](https://pypi.org/project/gridzen-developer-kit/0.6.0/); clean installation,
 CLI, stdio MCP and all six bundled Skill directories were verified. No extra MCP
 dependency installation is required.
 
@@ -99,7 +99,7 @@ gridzen setup --client all --project . --apply
 [Instructions / 使用说明](https://gridzen.ai/developers/harnesses.html) · [Compatibility / 测试记录](clients/compatibility.json)
 
 
-## Customer tasks and voluntary feedback (0.5.0)
+## Customer tasks and voluntary feedback (0.6.0)
 
 ```sh
 gridzen plan --country US --event onboarding --task signup_phone
@@ -113,4 +113,21 @@ gridzen feedback --input task-summary.json --share
 
 Phone intelligence is not OTP or ownership verification. Identity fixtures are not hosted document/liveness sessions. Payout matching remains a prototype with no real enabled route. Provider intake Skills are for provider operations/BYO, not a universal customer gate.
 
-Remote APIs count fixed task/capability/country/stage/error categories for 90 days, including technical failures. They store no arguments, identity data, credentials, raw chats or per-user tracking. Offline Skills/CLI never upload automatically. Optional feedback requires explicit consent, contains only fixed categories, and is unverified self-reported evidence. A SHA-256 hash of its random receipt ID is retained 30 days for within-day retry deduplication; the ID itself is not stored. Counters are capped at 10,000 requests/day and 500 summaries/day per database; missing history or offline usage is unknown. Requests are not unique customers.
+Remote APIs count fixed task/capability/country/stage/error categories for 90 days, including technical failures. They store no arguments, identity data, credentials, raw chats or per-user tracking. Local feedback is off by default. After explicit opt-in, CLI and stdio MCP automatically upload category-only events. Reading static Skills remains unobservable. Optional feedback requires explicit consent, contains only fixed categories, and is unverified self-reported evidence. A SHA-256 hash of its random receipt ID is retained 30 days for 30-day retry deduplication; the ID itself is not stored. Counters are capped at 10,000 requests/day and 500 summaries/day per database; missing history or offline usage is unknown. Requests are not unique customers.
+
+## 首次同意后的本地自动反馈（0.6.0）
+
+```sh
+gridzen telemetry status
+# Run only after reading and agreeing to category-only automatic sharing:
+gridzen telemetry enable --consent
+gridzen plan --country US --event onboarding --task signup_phone
+# Send one queued event per invocation after reconnecting:
+gridzen telemetry flush
+# Disable future uploads and clear the pending queue:
+gridzen telemetry disable
+```
+
+No network calls occur from local CLI/MCP when disabled. Consent is saved per OS user, not silently added to a project. Events contain only documented categories, ISO country, and a random per-event receipt, with no conversation, identity fields, credentials, user or project IDs. Feedback failures do not fail tool execution. The local queue is capped at 1,000 events/30 days; reconnect or `flush` retries one queued event, with a one-second HTTP timeout. Uploads are self-reported usage, not authenticated customers or completed verification. No background daemon is installed. Merely reading a Skill is unobservable; after consent, an Agent can explicitly record a fixed-category task summary.
+
+本地反馈默认关闭。用户先阅读说明并同意，再运行 `gridzen telemetry enable --consent`；此后 CLI 和本地 MCP 自动反馈分类事件。`gridzen telemetry disable` 立即关闭并清空待发送队列。断网最多保留 1000 条、30 天；恢复后每次调用或 flush 发送一条，无后台守护进程。只发送固定分类、国家和随机事件回执，不发送对话、身份资料、凭证或用户/项目标识。按服务器收到的日期统计，延迟上传不是该日实际发生的全部调用。仅阅读 Skill 无法自动观察，Agent 可在用户已经同意时主动 record 分类小结。它们仍是自报线索，不代表认证客户或真实核验完成。

{gridzen_developer_kit-0.5.0/src/gridzen_developer/web → gridzen_developer_kit-0.6.0/clients}/compatibility.json

@@ -1,5 +1,5 @@
 {
-  "version": "0.5.0",
+  "version": "0.6.0",
   "checked_at": "2026-10-09",
   "tools": 5,
   "skills": 6,

{gridzen_developer_kit-0.5.0 → gridzen_developer_kit-0.6.0}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "gridzen-developer-kit"
-version = "0.5.0"
+version = "0.6.0"
 readme = "README.md"
 license = "MIT"
 license-files = ["LICENSE", "NOTICE.md"]

{gridzen_developer_kit-0.5.0 → gridzen_developer_kit-0.6.0}/skills/gridzen-explain-verification/SKILL.md

@@ -22,3 +22,6 @@ Separate network/auth/schema/rate-limit errors from returned synthetic fixture o
 ## Optional task feedback
 
 Offline use is not tracked. Offer a category-only local task summary when the user wants to report a blocker: `gridzen feedback --task signup_phone --country US --requested-capability phone_intelligence --outcome blocked --blocker missing_workflow_step --skill gridzen-integrate-sandbox --output task-summary.json`. Choose the actual fixed categories; do not fabricate demand. Inspect the file with the user. Only an explicit user choice permits `gridzen feedback --input task-summary.json --share`; otherwise stop at the local file. Do not include prompts, emails, phone numbers, identity documents, account numbers, credentials or free text. Sharing is self-reported evidence and is not verified customer completion. No hidden upload, background call or new MCP tool is required.
+
+
+Local feedback in kit 0.6.0 is off by default. If the CLI is installed, `gridzen telemetry status` shows consent. Never enable it on the user's behalf without explicit agreement. When enabled, CLI/stdio calls automatically send fixed-category technical events; do not duplicate those events. For a Skill-only workflow, record one task summary after actual work with `gridzen telemetry record --skill gridzen-explain-verification --task <documented-task> --outcome <documented-outcome> --blocker <documented-reason>`, only if consent is already enabled. Omit raw text and personal data; choose unknown rather than infer completion. Do not install a dependency just to report telemetry. The user can disable and clear pending feedback with `gridzen telemetry disable`. Reading these instructions alone is not tracked.

{gridzen_developer_kit-0.5.0 → gridzen_developer_kit-0.6.0}/skills/gridzen-integrate-sandbox/SKILL.md

@@ -21,3 +21,6 @@ For signup phone checks use `phone_intelligence`; number type is not ownership o
 ## Optional task feedback
 
 Offline use is not tracked. Offer a category-only local task summary when the user wants to report a blocker: `gridzen feedback --task signup_phone --country US --requested-capability phone_intelligence --outcome blocked --blocker missing_workflow_step --skill gridzen-integrate-sandbox --output task-summary.json`. Choose the actual fixed categories; do not fabricate demand. Inspect the file with the user. Only an explicit user choice permits `gridzen feedback --input task-summary.json --share`; otherwise stop at the local file. Do not include prompts, emails, phone numbers, identity documents, account numbers, credentials or free text. Sharing is self-reported evidence and is not verified customer completion. No hidden upload, background call or new MCP tool is required.
+
+
+Local feedback in kit 0.6.0 is off by default. If the CLI is installed, `gridzen telemetry status` shows consent. Never enable it on the user's behalf without explicit agreement. When enabled, CLI/stdio calls automatically send fixed-category technical events; do not duplicate those events. For a Skill-only workflow, record one task summary after actual work with `gridzen telemetry record --skill gridzen-integrate-sandbox --task <documented-task> --outcome <documented-outcome> --blocker <documented-reason>`, only if consent is already enabled. Omit raw text and personal data; choose unknown rather than infer completion. Do not install a dependency just to report telemetry. The user can disable and clear pending feedback with `gridzen telemetry disable`. Reading these instructions alone is not tracked.

{gridzen_developer_kit-0.5.0 → gridzen_developer_kit-0.6.0}/skills/gridzen-integrate-sandbox/references/api.md

@@ -23,4 +23,4 @@ assert result['status'] == 'inconclusive'
 Unexpected fields are rejected (422); unknown country/capability/fixture is rejected (400). Network errors and HTTP failures are integration failures, not verification outcomes. Do not retry indefinitely. Local core and local MCP work offline after installation; Python HTTP SDK sends only documented fixture parameters to Gridzen.
 
 
-Task-specific planning (kit 0.5.0): `plan_verification(country="US", event="onboarding", task="signup_phone", stage="prototype")`. Identity task: `identity_onboarding`. Task and event must agree. Fixtures accept `phone_intelligence` and `identity_verification` in addition to existing research capability names. All fixture outcomes remain simulated and verified=false. Production planning always reports NO_LIVE_ROUTE_ENABLED. No extra MCP tool is added.
+Task-specific planning (kit 0.6.0): `plan_verification(country="US", event="onboarding", task="signup_phone", stage="prototype")`. Identity task: `identity_onboarding`. Task and event must agree. Fixtures accept `phone_intelligence` and `identity_verification` in addition to existing research capability names. All fixture outcomes remain simulated and verified=false. Production planning always reports NO_LIVE_ROUTE_ENABLED. No extra MCP tool is added.

{gridzen_developer_kit-0.5.0 → gridzen_developer_kit-0.6.0}/skills/gridzen-select-verification/SKILL.md

@@ -19,3 +19,6 @@ Select the proof for the customer task before choosing a provider. Use `plan_ver
 ## Optional task feedback
 
 Offline use is not tracked. Offer a category-only local task summary when the user wants to report a blocker: `gridzen feedback --task signup_phone --country US --requested-capability phone_intelligence --outcome blocked --blocker missing_workflow_step --skill gridzen-integrate-sandbox --output task-summary.json`. Choose the actual fixed categories; do not fabricate demand. Inspect the file with the user. Only an explicit user choice permits `gridzen feedback --input task-summary.json --share`; otherwise stop at the local file. Do not include prompts, emails, phone numbers, identity documents, account numbers, credentials or free text. Sharing is self-reported evidence and is not verified customer completion. No hidden upload, background call or new MCP tool is required.
+
+
+Local feedback in kit 0.6.0 is off by default. If the CLI is installed, `gridzen telemetry status` shows consent. Never enable it on the user's behalf without explicit agreement. When enabled, CLI/stdio calls automatically send fixed-category technical events; do not duplicate those events. For a Skill-only workflow, record one task summary after actual work with `gridzen telemetry record --skill gridzen-select-verification --task <documented-task> --outcome <documented-outcome> --blocker <documented-reason>`, only if consent is already enabled. Omit raw text and personal data; choose unknown rather than infer completion. Do not install a dependency just to report telemetry. The user can disable and clear pending feedback with `gridzen telemetry disable`. Reading these instructions alone is not tracked.

{gridzen_developer_kit-0.5.0 → gridzen_developer_kit-0.6.0}/skills/mexico-pilot-scoper/SKILL.md

@@ -22,3 +22,6 @@ Mexico payout scope remains a bounded draft. No real bank matching route is enab
 ## Optional task feedback
 
 Offline use is not tracked. Offer a category-only local task summary when the user wants to report a blocker: `gridzen feedback --task signup_phone --country US --requested-capability phone_intelligence --outcome blocked --blocker missing_workflow_step --skill gridzen-integrate-sandbox --output task-summary.json`. Choose the actual fixed categories; do not fabricate demand. Inspect the file with the user. Only an explicit user choice permits `gridzen feedback --input task-summary.json --share`; otherwise stop at the local file. Do not include prompts, emails, phone numbers, identity documents, account numbers, credentials or free text. Sharing is self-reported evidence and is not verified customer completion. No hidden upload, background call or new MCP tool is required.
+
+
+Local feedback in kit 0.6.0 is off by default. If the CLI is installed, `gridzen telemetry status` shows consent. Never enable it on the user's behalf without explicit agreement. When enabled, CLI/stdio calls automatically send fixed-category technical events; do not duplicate those events. For a Skill-only workflow, record one task summary after actual work with `gridzen telemetry record --skill mexico-pilot-scoper --task <documented-task> --outcome <documented-outcome> --blocker <documented-reason>`, only if consent is already enabled. Omit raw text and personal data; choose unknown rather than infer completion. Do not install a dependency just to report telemetry. The user can disable and clear pending feedback with `gridzen telemetry disable`. Reading these instructions alone is not tracked.

{gridzen_developer_kit-0.5.0 → gridzen_developer_kit-0.6.0}/skills/payout-policy-designer/SKILL.md

@@ -22,3 +22,6 @@ Payout account matching currently has no real enabled Gridzen route. Keep output
 ## Optional task feedback
 
 Offline use is not tracked. Offer a category-only local task summary when the user wants to report a blocker: `gridzen feedback --task signup_phone --country US --requested-capability phone_intelligence --outcome blocked --blocker missing_workflow_step --skill gridzen-integrate-sandbox --output task-summary.json`. Choose the actual fixed categories; do not fabricate demand. Inspect the file with the user. Only an explicit user choice permits `gridzen feedback --input task-summary.json --share`; otherwise stop at the local file. Do not include prompts, emails, phone numbers, identity documents, account numbers, credentials or free text. Sharing is self-reported evidence and is not verified customer completion. No hidden upload, background call or new MCP tool is required.
+
+
+Local feedback in kit 0.6.0 is off by default. If the CLI is installed, `gridzen telemetry status` shows consent. Never enable it on the user's behalf without explicit agreement. When enabled, CLI/stdio calls automatically send fixed-category technical events; do not duplicate those events. For a Skill-only workflow, record one task summary after actual work with `gridzen telemetry record --skill payout-policy-designer --task <documented-task> --outcome <documented-outcome> --blocker <documented-reason>`, only if consent is already enabled. Omit raw text and personal data; choose unknown rather than infer completion. Do not install a dependency just to report telemetry. The user can disable and clear pending feedback with `gridzen telemetry disable`. Reading these instructions alone is not tracked.

{gridzen_developer_kit-0.5.0 → gridzen_developer_kit-0.6.0}/skills/provider-rights-readiness/SKILL.md

@@ -22,3 +22,6 @@ This is a provider-operation or bring-your-own-provider task. Do not impose this
 ## Optional task feedback
 
 Offline use is not tracked. Offer a category-only local task summary when the user wants to report a blocker: `gridzen feedback --task signup_phone --country US --requested-capability phone_intelligence --outcome blocked --blocker missing_workflow_step --skill gridzen-integrate-sandbox --output task-summary.json`. Choose the actual fixed categories; do not fabricate demand. Inspect the file with the user. Only an explicit user choice permits `gridzen feedback --input task-summary.json --share`; otherwise stop at the local file. Do not include prompts, emails, phone numbers, identity documents, account numbers, credentials or free text. Sharing is self-reported evidence and is not verified customer completion. No hidden upload, background call or new MCP tool is required.
+
+
+Local feedback in kit 0.6.0 is off by default. If the CLI is installed, `gridzen telemetry status` shows consent. Never enable it on the user's behalf without explicit agreement. When enabled, CLI/stdio calls automatically send fixed-category technical events; do not duplicate those events. For a Skill-only workflow, record one task summary after actual work with `gridzen telemetry record --skill provider-rights-readiness --task <documented-task> --outcome <documented-outcome> --blocker <documented-reason>`, only if consent is already enabled. Omit raw text and personal data; choose unknown rather than infer completion. Do not install a dependency just to report telemetry. The user can disable and clear pending feedback with `gridzen telemetry disable`. Reading these instructions alone is not tracked.

{gridzen_developer_kit-0.5.0 → gridzen_developer_kit-0.6.0}/src/gridzen_developer/__init__.py

@@ -1,2 +1,2 @@
 """Gridzen developer tools. Research and simulation are never live verification."""
-__version__ = '0.5.0'
+__version__ = '0.6.0'

{gridzen_developer_kit-0.5.0 → gridzen_developer_kit-0.6.0}/src/gridzen_developer/cli.py

@@ -18,7 +18,26 @@ def main():
  from .setup import CLIENTS
  x=s.add_parser('setup',help='Prepare six Skills and MCP config for a harness; dry-run until --apply')
  x.add_argument('--client',choices=['all',*CLIENTS],required=True);x.add_argument('--project',default='.');x.add_argument('--apply',action='store_true')
+ x=s.add_parser('telemetry',help='Opt-in local category feedback; no chat, personal fields or credentials')
+ x.add_argument('action',choices=['enable','disable','status','flush','record'])
+ x.add_argument('--consent',action='store_true',help='Explicitly consent to automatic category uploads')
+ x.add_argument('--task',choices=sorted(TASKS),default='unknown');x.add_argument('--skill',choices=sorted(SKILLS),default='unknown')
+ x.add_argument('--outcome',choices=sorted(OUTCOMES),default='unknown');x.add_argument('--blocker',choices=sorted(REASONS),default='unknown')
  a=p.parse_args()
+ if a.command=='telemetry':
+  from . import telemetry
+  try:
+   if a.action=='enable':
+    if not a.consent:p.error('Enabling automatic feedback requires --consent. Fixed categories only; disable anytime.')
+    result=telemetry.configure(True)
+   elif a.action=='disable':result=telemetry.configure(False)
+   elif a.action=='status':result=telemetry.status()
+   elif a.action=='flush':result=telemetry.flush()
+   else:
+    telemetry.observe({'operation':'task_summary','task':a.task,'skill':a.skill,'mode':'local_draft','outcome':a.outcome,'reason':a.blocker})
+    result=telemetry.status()
+  except (OSError,ValueError) as exc:p.error(type(exc).__name__)
+  print(json.dumps(result,indent=2));return
  if a.command=='feedback':
   from .product_feedback import validate_summary
   try:
@@ -43,6 +62,12 @@ def main():
   print(json.dumps(list_skills(),ensure_ascii=False,indent=2));return
  try:
   result={'coverage':lambda:core.coverage(a.country,a.capability),'plan':lambda:core.plan(a.country,a.event,a.task,a.stage),'simulate':lambda:core.simulate(a.country,a.capability,a.scenario),'explain':lambda:core.explain(a.reason_code)}[a.command]()
- except ValueError as e:p.error(str(e))
+ except ValueError as e:
+  from .telemetry import observe
+  observe({'operation':{'coverage':'get_coverage','plan':'plan_verification','simulate':'create_sandbox_verification','explain':'explain_result'}[a.command],'task':getattr(a,'task',None),'country':getattr(a,'country',None),'outcome':'integration_error','reason':'schema_error'})
+  p.error(str(e))
+ from .telemetry import observe
+ blocked=a.command=='plan' and (a.stage=='production' or result.get('task_match_status') in ('unsupported','needs_clarification'))
+ observe({'operation':{'coverage':'get_coverage','plan':'plan_verification','simulate':'create_sandbox_verification','explain':'explain_result'}[a.command],'task':getattr(a,'task',None),'country':getattr(a,'country',None),'capability':result.get('capability',getattr(a,'capability',None)),'stage':getattr(a,'stage','prototype'),'mode':'research' if a.command in ('coverage','plan') else 'synthetic_fixture','outcome':'capability_unavailable' if blocked else 'technical_success','reason':'production_disabled' if blocked and a.stage=='production' else 'unavailable_route' if blocked else 'none'})
  print(json.dumps(result,ensure_ascii=False,indent=2))
 if __name__=='__main__':main()

{gridzen_developer_kit-0.5.0 → gridzen_developer_kit-0.6.0}/src/gridzen_developer/mcp.py

@@ -27,7 +27,7 @@ class GridzenMCP(MCPServer):
   for tool in tools:tool.input_schema['additionalProperties']=False
   return tools
 
-server=GridzenMCP('Gridzen Developer Kit',version='0.5.0',website_url='https://gridzen.ai/developers/',middleware=[strict_arguments],instructions='Research and synthetic fixtures only. No live providers are enabled. Preserve verified=false; never use a sandbox result to approve a real person or payment.')
+server=GridzenMCP('Gridzen Developer Kit',version='0.6.0',website_url='https://gridzen.ai/developers/',middleware=[strict_arguments],instructions='Research and synthetic fixtures only. No live providers are enabled. Preserve verified=false; never use a sandbox result to approve a real person or payment.')
 read=ToolAnnotations(readOnlyHint=True,destructiveHint=False,idempotentHint=True,openWorldHint=False)
 @server.tool(annotations=read,structured_output=True)
 def get_coverage(country: str | None=None, capability: str | None=None) -> dict[str, Any]:
@@ -50,5 +50,27 @@ def explain_result(reason_code: str) -> dict[str, Any]:
  """Explain a sandbox reason code; missing data and timeout remain inconclusive."""
  return core.explain(reason_code)
 TOOLS={fn.__name__:fn for fn in (get_coverage,plan_verification,create_sandbox_verification,get_sandbox_verification,explain_result)}
-def main():server.run(transport='stdio')
+async def local_feedback(ctx, call_next):
+ from .telemetry import observe
+ import asyncio
+ params=ctx.params if isinstance(ctx.params,dict) else {}
+ args=params.get('arguments') if isinstance(params.get('arguments'),dict) else {}
+ dims={'operation':params.get('name'),'task':args.get('task'),'country':args.get('country'),'capability':args.get('capability'),'stage':args.get('stage','prototype'),'mode':'research' if params.get('name') in ('get_coverage','plan_verification') else 'synthetic_fixture'}
+ try: result=await call_next(ctx)
+ except Exception:
+  if ctx.method=='tools/call':await asyncio.to_thread(observe,{**dims,'outcome':'integration_error','reason':'tool_error'})
+  raise
+ if ctx.method=='tools/call':
+  content=getattr(result,'structured_content',None) or {}
+  blocked=content.get('requested_stage')=='production' or content.get('task_match_status') in ('unsupported','needs_clarification')
+  failed=getattr(result,'is_error',False)
+  await asyncio.to_thread(observe,{**dims,'capability':content.get('capability',dims['capability']),'outcome':'integration_error' if failed else 'capability_unavailable' if blocked else 'technical_success','reason':'tool_error' if failed else 'production_disabled' if blocked and content.get('requested_stage')=='production' else 'unavailable_route' if blocked else 'none'})
+ return result
+
+def main():
+ from .telemetry import enabled
+ if enabled():
+  read.open_world_hint=True
+ server.middleware.insert(0,local_feedback)
+ server.run(transport='stdio')
 if __name__=='__main__':main()

{gridzen_developer_kit-0.5.0 → gridzen_developer_kit-0.6.0}/src/gridzen_developer/product_feedback.py

@@ -15,7 +15,7 @@ TASKS = frozenset(('signup_phone','identity_onboarding','payout_account','phone_
 CAPABILITIES = frozenset(('official_identity','commercial_identity','consented_eid','bank_account_match','consented_bank','cardholder_name','phone_identity','phone_intelligence','identity_verification','otp','business_kyb','batch','webhook_recovery','policy_design','provider_rights','other','unknown'))
 SKILLS = frozenset(('gridzen-select-verification','gridzen-integrate-sandbox','gridzen-explain-verification','provider-rights-readiness','payout-policy-designer','mexico-pilot-scoper','unknown'))
 OPERATIONS = frozenset(('get_coverage','plan_verification','create_sandbox_verification','get_sandbox_verification','explain_result','merchant_create','merchant_read','merchant_simulate','merchant_refresh','merchant_capabilities','task_summary','transport_rejected','other'))
-SOURCES = frozenset(('internal_test','self_reported_test','unknown','authenticated_merchant_sandbox','self_reported_feedback','authenticated_feedback'))
+SOURCES = frozenset(('internal_test','self_reported_test','unknown','authenticated_merchant_sandbox','self_reported_feedback','authenticated_feedback','self_reported_local_usage'))
 MODES = frozenset(('research','synthetic_fixture','merchant_sandbox','local_draft','unknown'))
 OUTCOMES = frozenset(('technical_success','integration_error','capability_unavailable','completed_prototype','blocked','abandoned','unknown'))
 REASONS = frozenset(('none','auth_error','schema_error','rate_limit','timeout','transport_error','tool_error','unknown_country','unknown_capability','unavailable_route','coverage_gap','missing_workflow_step','price_unknown','production_disabled','already_solved_elsewhere','idempotency_conflict','record_not_found','other','unknown'))
@@ -57,7 +57,7 @@ def write(path, values, receipt=None, day=None):
             feedback=db.execute("SELECT COALESCE(SUM(count),0) FROM product_events WHERE day=? AND operation='task_summary'",(day.isoformat(),)).fetchone()[0]
             if feedback>=500:return False
             digest=hashlib.sha256(receipt.encode()).hexdigest()
-            if db.execute('SELECT 1 FROM feedback_receipts WHERE day=? AND hash=?',(day.isoformat(),digest)).fetchone():return True
+            if db.execute('SELECT 1 FROM feedback_receipts WHERE hash=?',(digest,)).fetchone():return True
             db.execute('INSERT INTO feedback_receipts VALUES (?,?)',(day.isoformat(),digest))
         columns='day,'+','.join(FIELDS)
         values=(day.isoformat(),*(safe[key] for key in FIELDS))
@@ -93,3 +93,16 @@ def accept_summary(path, value, source='self_reported_feedback'):
 
 def http_reason(status):
     return {400:'schema_error',401:'auth_error',403:'auth_error',404:'record_not_found',409:'idempotency_conflict',413:'schema_error',422:'schema_error',429:'rate_limit',502:'unavailable_route',503:'unavailable_route',504:'timeout'}.get(status,'tool_error' if status>=400 else 'none')
+
+
+def accept_usage(path, value):
+    if not isinstance(value,dict) or set(value)!={'schema_version','event_id','consent_to_share','dimensions'} or type(value['schema_version']) is not int or value['schema_version']!=1 or value['consent_to_share'] is not True:
+        raise ValueError('INVALID_USAGE_EVENT')
+    if not isinstance(value['event_id'],str) or len(value['event_id'])!=36: raise ValueError('INVALID_USAGE_EVENT')
+    try: UUID(value['event_id'])
+    except ValueError: raise ValueError('INVALID_USAGE_EVENT') from None
+    dimensions=value['dimensions']
+    if not isinstance(dimensions,dict) or set(dimensions)!=set(FIELDS) or safe_dimensions(dimensions)!=dimensions or dimensions['source']!='self_reported_local_usage': raise ValueError('INVALID_USAGE_EVENT')
+    if dimensions['operation'] not in OPERATIONS-{'merchant_create','merchant_read','merchant_simulate','merchant_refresh','merchant_capabilities','transport_rejected'}: raise ValueError('INVALID_USAGE_EVENT')
+    if dimensions['mode']=='merchant_sandbox': raise ValueError('INVALID_USAGE_EVENT')
+    return write(path,dimensions,receipt='local:'+value['event_id'])

{gridzen_developer_kit-0.5.0 → gridzen_developer_kit-0.6.0}/src/gridzen_developer/server.py

@@ -18,7 +18,7 @@ mcp_app=make_transport()
 async def lifespan(_app):
  async with mcp_app.router.lifespan_context(mcp_app):yield
 
-app=FastAPI(title='Gridzen Developer Sandbox',version='0.5.0',lifespan=lifespan,docs_url=None,redoc_url=None,openapi_url='/api/openapi.json',servers=[{'url':'/developers'}])
+app=FastAPI(title='Gridzen Developer Sandbox',version='0.6.0',lifespan=lifespan,docs_url=None,redoc_url=None,openapi_url='/api/openapi.json',servers=[{'url':'/developers'}])
 app.add_middleware(MCPBounds)
 app.add_middleware(CallSourceMiddleware)
 class PlanRequest(BaseModel):
@@ -67,10 +67,10 @@ async def bounds(request:Request,call_next):
  response.headers['Cache-Control']='no-store' if request.url.path.startswith(('/api','/mcp')) else 'public, max-age=60'
  return response
 @app.get('/health')
-def health():return {'status':'ok','mode':'sandbox','live_routes':0,'version':'0.5.0','mcp_transport':'streamable-http'}
+def health():return {'status':'ok','mode':'sandbox','live_routes':0,'version':'0.6.0','mcp_transport':'streamable-http'}
 @app.get('/server-card.json')
 async def server_card():
- return {'serverInfo':{'name':'Gridzen Verification','version':'0.5.0'},'authentication':{'required':False},'tools':[tool.model_dump(by_alias=True,exclude_none=True) for tool in await mcp_server.list_tools()],'resources':[],'prompts':[]}
+ return {'serverInfo':{'name':'Gridzen Verification','version':'0.6.0'},'authentication':{'required':False},'tools':[tool.model_dump(by_alias=True,exclude_none=True) for tool in await mcp_server.list_tools()],'resources':[],'prompts':[]}
 @app.get('/api/coverage')
 def coverage(country:str|None=Query(None,max_length=2),capability:str|None=Query(None,max_length=40)):return core.coverage(country,capability)
 @app.post('/api/plan')
@@ -94,5 +94,14 @@ async def feedback(request:Request):
  if not accepted:return JSONResponse({'error':'FEEDBACK_TEMPORARILY_UNAVAILABLE'},status_code=503)
  return {'accepted':True,'evidence':'self_reported','authenticated_customer':False,'retention':'aggregate 90 days; receipt hash 30 days','duplicate_receipts':'deduplicated within UTC day'}
 
+@app.post('/api/usage-events')
+async def local_usage(request:Request):
+ from .product_feedback import accept_usage
+ try: accepted=accept_usage(os.environ.get('GRIDZEN_MCP_EVENTS_DB'),json.loads(await request.body()))
+ except ValueError: return JSONResponse({'error':'INVALID_USAGE_EVENT'},status_code=422)
+ except (OSError,sqlite3.Error): accepted=False
+ if not accepted: return JSONResponse({'error':'USAGE_TEMPORARILY_UNAVAILABLE'},status_code=503)
+ return {'accepted':True,'evidence':'self_reported_local_usage','authenticated_customer':False}
+
 app.router.routes.extend(mcp_app.routes)
 app.mount('/',StaticFiles(directory=Path(__file__).parent/'web',html=True),name='developer-web')