From 6292f011884dc4a6cf253609dda4f27141423335 Mon Sep 17 00:00:00 2001 From: elkimek <36666630+elkimek@users.noreply.github.com> Date: Thu, 14 May 2026 13:03:20 +0200 Subject: [PATCH 1/2] Add Agent Skills discovery index MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Publish /.well-known/agent-skills/index.json (Agent Skills Discovery RFC v0.2.0) with one skill — getbased-health-data — that teaches agents how to connect to and responsibly query a user's getbased health data via the MCP server. Co-Authored-By: Claude Opus 4.7 (1M context) --- .../getbased-health-data/SKILL.md | 45 +++++++++++++++++++ .well-known/agent-skills/index.json | 12 +++++ 2 files changed, 57 insertions(+) create mode 100644 .well-known/agent-skills/getbased-health-data/SKILL.md create mode 100644 .well-known/agent-skills/index.json diff --git a/.well-known/agent-skills/getbased-health-data/SKILL.md b/.well-known/agent-skills/getbased-health-data/SKILL.md new file mode 100644 index 0000000..5671eef --- /dev/null +++ b/.well-known/agent-skills/getbased-health-data/SKILL.md @@ -0,0 +1,45 @@ +--- +name: getbased-health-data +description: Connect to a user's getbased health data and query their blood work, biomarkers, genome, wearables, and personal research library via the getbased MCP server. Use when a user asks about their lab results, health trends, or wants AI reasoning grounded in their own health data. +--- + +# getbased Health Data + +[getbased](https://getbased.health) is a free, open-source Personal Health Intelligence app that runs in the user's browser. It connects five lenses on a person's biology — labs, genome, body, lifestyle, and environment — and exposes a read-only summary to AI agents through the [`getbased-mcp`](https://github.com/elkimek/getbased-agents/tree/main/packages/mcp) server. + +The user's raw data and encryption keys never leave their device. The MCP server reads the same pre-built context text the in-app AI chat uses — not the underlying database. + +## When to use this skill + +Use this skill when the user wants you to reason about *their own* health data: "how's my vitamin D", "what markers are out of range", "summarize my latest blood work", "what does my research library say about X". Do not use it for general medical questions unconnected to the user's data. + +## Setup + +The MCP server is installed and run locally by the user — there is no hosted endpoint. + +1. Install: `pip install getbased-mcp` (or the `getbased-agent-stack` bundle). +2. The user enables **Settings > Data > Messenger Access** in the getbased app and copies the read-only token. +3. The token is provided to the MCP server via the `GETBASED_TOKEN` environment variable. It grants access to lab-context text only, no raw data, and is revocable. +4. Add `getbased` to the MCP client config pointing at the `getbased-mcp` command. + +Full configuration details: https://app.getbased.health/docs/guide/agent-access + +## Tools + +| Tool | Use it for | +|---|---| +| `getbased_lab_context` | Full lab summary — biomarkers, ranges, trends, context cards, supplements, goals | +| `getbased_section` | One section (e.g. hormones, lipids) instead of the full dump — token-efficient | +| `getbased_list_profiles` | List profiles by name and ID; pass `profile` to any tool to target one | +| `knowledge_search` | Semantic search over the user's own research library (requires the optional RAG server) | +| `knowledge_list_libraries` / `knowledge_activate_library` / `knowledge_stats` | Manage which research library `knowledge_search` targets | +| `getbased_lens_config` | Show the Custom Knowledge Source endpoint configuration | + +Start with `getbased_section()` (no args) to see the section index, then pull only what you need. Knowledge tools degrade gracefully — if the RAG server is down, the lab tools still work. + +## Responsible use + +- You are not a clinician. Interpret data, surface patterns and out-of-range markers, and cite reference/optimal ranges — but recommend the user confirm anything actionable with a healthcare provider. +- Ground claims in the user's actual data and, where available, their research library via `knowledge_search`. Prefer cited passages over training-data recall for health claims. +- The context is a summary, not raw records. Don't infer precision the data doesn't support. +- Respect multi-profile boundaries — never mix data across profiles unless the user asks. diff --git a/.well-known/agent-skills/index.json b/.well-known/agent-skills/index.json new file mode 100644 index 0000000..432db96 --- /dev/null +++ b/.well-known/agent-skills/index.json @@ -0,0 +1,12 @@ +{ + "$schema": "https://schemas.agentskills.io/discovery/0.2.0/schema.json", + "skills": [ + { + "name": "getbased-health-data", + "type": "skill-md", + "description": "Connect to a user's getbased health data and query their blood work, biomarkers, genome, wearables, and personal research library via the getbased MCP server. Use when a user asks about their lab results, health trends, or wants AI reasoning grounded in their own health data.", + "url": "/.well-known/agent-skills/getbased-health-data/SKILL.md", + "digest": "sha256:38b639d200a4bff406f3d5dfc83ed5246a59630806c42cb0cc31588776ada99c" + } + ] +} From f20b40bf77b1ed44403fcb4ff58e3563f5a8d808 Mon Sep 17 00:00:00 2001 From: elkimek <36666630+elkimek@users.noreply.github.com> Date: Thu, 14 May 2026 13:10:53 +0200 Subject: [PATCH 2/2] Correct MCP metadata against getbased-mcp v0.2.5 Audited the discovery files against the getbased-agents monorepo source: - Add the getbased_wearables_series tool (9 tools total, was 8) to both the MCP server card and the SKILL.md tool table - Fix license: AGPL-3.0-or-later (was GPL-3.0-only) - Bump version to 0.2.5 - Regenerate the SKILL.md digest in the skills index Co-Authored-By: Claude Opus 4.7 (1M context) --- .well-known/agent-skills/getbased-health-data/SKILL.md | 1 + .well-known/agent-skills/index.json | 2 +- .well-known/mcp.json | 5 +++-- 3 files changed, 5 insertions(+), 3 deletions(-) diff --git a/.well-known/agent-skills/getbased-health-data/SKILL.md b/.well-known/agent-skills/getbased-health-data/SKILL.md index 5671eef..e62cbf8 100644 --- a/.well-known/agent-skills/getbased-health-data/SKILL.md +++ b/.well-known/agent-skills/getbased-health-data/SKILL.md @@ -30,6 +30,7 @@ Full configuration details: https://app.getbased.health/docs/guide/agent-access |---|---| | `getbased_lab_context` | Full lab summary — biomarkers, ranges, trends, context cards, supplements, goals | | `getbased_section` | One section (e.g. hormones, lipids) instead of the full dump — token-efficient | +| `getbased_wearables_series` | Daily wearable time-series (HRV, resting HR, sleep score, readiness, steps, weight…) over the 7/30/90-day window the user opted into | | `getbased_list_profiles` | List profiles by name and ID; pass `profile` to any tool to target one | | `knowledge_search` | Semantic search over the user's own research library (requires the optional RAG server) | | `knowledge_list_libraries` / `knowledge_activate_library` / `knowledge_stats` | Manage which research library `knowledge_search` targets | diff --git a/.well-known/agent-skills/index.json b/.well-known/agent-skills/index.json index 432db96..1b363f6 100644 --- a/.well-known/agent-skills/index.json +++ b/.well-known/agent-skills/index.json @@ -6,7 +6,7 @@ "type": "skill-md", "description": "Connect to a user's getbased health data and query their blood work, biomarkers, genome, wearables, and personal research library via the getbased MCP server. Use when a user asks about their lab results, health trends, or wants AI reasoning grounded in their own health data.", "url": "/.well-known/agent-skills/getbased-health-data/SKILL.md", - "digest": "sha256:38b639d200a4bff406f3d5dfc83ed5246a59630806c42cb0cc31588776ada99c" + "digest": "sha256:21e2c05fab6d810842d48562af259d5dde9cd005152ae4463aa4cbb0c189fc3d" } ] } diff --git a/.well-known/mcp.json b/.well-known/mcp.json index 58d5774..c660961 100644 --- a/.well-known/mcp.json +++ b/.well-known/mcp.json @@ -1,13 +1,13 @@ { "name": "getbased", "description": "MCP server that exposes your getbased health data — blood work context, biomarker sections, and an optional RAG knowledge base — as tools for any MCP-compatible agent. Your data stays on your device; the server reads a read-only context summary, not raw records.", - "version": "0.2.0", + "version": "0.2.5", "transport": "stdio", "hosting": "local", "homepage": "https://getbased.health", "documentation": "https://app.getbased.health/docs/guide/agent-access", "repository": "https://github.com/elkimek/getbased-agents/tree/main/packages/mcp", - "license": "GPL-3.0-only", + "license": "AGPL-3.0-or-later", "install": { "pypi": "getbased-mcp", "command": "getbased-mcp", @@ -26,6 +26,7 @@ "tools": [ { "name": "getbased_lab_context", "description": "Full lab summary — biomarkers, ranges, trends, context cards, supplements, goals. Pass a profile to target a specific profile." }, { "name": "getbased_section", "description": "Get a specific section (e.g. hormones, lipids) or list all available sections. Token-efficient alternative to the full dump." }, + { "name": "getbased_wearables_series", "description": "Read the daily wearable time-series (HRV, resting HR, sleep score, readiness, steps, weight, etc.) over the 7/30/90-day window the user opted into pushing." }, { "name": "getbased_list_profiles", "description": "List available profiles by name and ID." }, { "name": "knowledge_search", "description": "Semantic search across the active library of your RAG knowledge base. Requires the getbased-rag server; degrades gracefully if unavailable." }, { "name": "knowledge_list_libraries", "description": "List all knowledge-base libraries and show which is active." },