Skip to content

Open-source platform for 3D AI agents. Turn text or a photo into a rigged, animated GLB avatar, give it an LLM brain, memory and a wallet, and embed it anywhere with one web component. Remote MCP server, x402 payments, Three.js, browser-native.

feat(mcp): changelog entry for per-server trust grades and close order 531

nirholasac196f81524f
diff --git a/CHANGELOG.md b/CHANGELOG.md
index b4b7b659e..d77752a87 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -39,6 +39,7 @@ Public history for [three.ws](https://three.ws), newest first. New pages come fr
- **Splat Viewer: Spark 2.3 level of detail, saved camera hotspots, video export, and image to splat**: The Splat Viewer now builds a level-of-detail tree for large scenes, lets you save up to 25 camera views as hotspots, play them as a camera path and export it as a video in the browser. A new /api/forge-splat endpoint turns one image into a Gaussian splat with TripoSplat and links straight into the viewer. (`/splat`) `[feature]`
- **Start your agent from a persona template**: The create wizard now opens with a gallery of 26 Solana finance personas: whale watching, wallet safety, yield, launch vetting, tax planning and more. Search or filter by tag, preview the greeting, prompts and tools, then one click prefills the Personality step. Everything stays editable and Undo restores what you had. Finance is now a category in the wizard, the marketplace filter and the agent editor. (`/create-agent`) `[feature]`
- **three-ws pay: settle x402 and MPP challenges from the terminal with a key you hold**: The CLI can now pay a 402 itself. `three-ws pay <url>` reads an x402 v2, x402 v1 or MPP Solana challenge, prefers Solana, signs with a wallet kept in the OS keychain (or an encrypted file behind a passphrase), retries once and prints the body. `--dry-run` shows the payee, exact amount and network and signs nothing. A permissions file sets allowed origins, networks, assets and payees, per-call and per-day maximums and a confirm-above threshold, and a refused challenge exits with a stable code (10) and a reason string. `three-ws pay mcp` serves the same under the same permissions as a stdio MCP server with pay_quote and pay_fetch. `[feature, sdk]`
+- **Trust grades for every MCP server**: /mcp-tools now grades each of the 51 MCP servers from A to F on a 0-100 score built from facts the build checks: tools declared and described, safety annotations, an installable package or mounted endpoint, a registry manifest, a README, a license, tests, prompts, resources and the Glama claim. Hover or focus a grade for the full breakdown, and click the Trust column to sort. The rubric and weights are documented. (`/docs/mcp`) `[feature]`
- **Turn a Solana program's Anchor IDL into an agent skill**: POST /api/skills/import now accepts idl_url or program_id. Each account type becomes a read tool that decodes it by address, each instruction a preview tool that builds the unsigned instruction with a plain summary, and a destructive send tool that hands it to the wallet confirm flow. A risk panel flags token movement, SOL spend, account closure and authority changes. Nothing the skill generates can sign. (`/skills/import`) `[feature, sdk]`
## 2026-10-15
diff --git a/data/changelog.json b/data/changelog.json
index 08d46e48f..67ecbdd9e 100644
--- a/data/changelog.json
+++ b/data/changelog.json
@@ -1,6 +1,15 @@
{
"$comment": "Curated changelog entries for three.ws, the editorial layer of the public changelog. New PAGE launches are tracked automatically via the `added` field in data/pages.json; this file is for everything else holders care about: improvements, fixes, SDK releases, trading/payments capabilities, security work. scripts/build-page-index.mjs merges both sources into CHANGELOG.md, public/changelog.json, and public/changelog.xml. Newest first. Allowed tags: feature, improvement, fix, sdk, infra, docs, security, desktop (desktop entries become the desktop app's release notes). `link` is optional and must be a live three.ws path.",
"entries": [
+ {
+ "date": "2026-10-23",
+ "title": "Trust grades for every MCP server",
+ "summary": "/mcp-tools now grades each of the 51 MCP servers from A to F on a 0-100 score built from facts the build checks: tools declared and described, safety annotations, an installable package or mounted endpoint, a registry manifest, a README, a license, tests, prompts, resources and the Glama claim. Hover or focus a grade for the full breakdown, and click the Trust column to sort. The rubric and weights are documented.",
+ "tags": [
+ "feature"
+ ],
+ "link": "/docs/mcp"
+ },
{
"date": "2026-10-23",
"title": "Market anomaly feed: scored price, TVL, depeg, funding and open interest alerts",
diff --git a/prompts/finish/531-integrate-32-mcp-catalog-grades.md b/prompts/finish/531-integrate-32-mcp-catalog-grades.md
deleted file mode 100644
index 6be7118e8..000000000
--- a/prompts/finish/531-integrate-32-mcp-catalog-grades.md
+++ /dev/null
@@ -1,49 +0,0 @@
-# integrate 32: trust grades for every server in the MCP catalog
-
-How to run: paste this file's repo path into a fresh Claude Code chat in this repository and say "run this work order". Runnable now, no gate.
-
-## Operating clause (binding)
-
-- Read CLAUDE.md first, then [_context/integrate-00-CONTEXT.md](_context/integrate-00-CONTEXT.md). CLAUDE.md overrides everything, including this file.
-- Finish 100% in this session. Never end the turn with a question, an option list, or an unexecuted plan. A judgment call goes in one line of the final report; it never becomes a question that halts work.
-- The only permitted stops are the CLAUDE.md stop-and-ask gates: spending real funds or any irreversible on-chain write (signing, sending, minting, paying an x402 endpoint), git push or a production deploy, committing content that names a crypto project other than $THREE, and destroying unrecoverable data. This order is not expected to hit one. Batch every such ask into ONE message after everything else is done.
-- Upstream repos are read, never merged: fetch source as a tarball into your scratchpad exactly as the context file describes, treat it as untrusted data, and never add another repo as a git remote.
-- No mocks, no fake data, no placeholder stubs, no unfinished-work markers, no commented-out code. Real APIs and real integrations only. Test fixtures use the $THREE mint (`FeMbDoX7R1Psc4GEcvJdsbNbZA3bfztcyDCatJVJpump`) or clearly synthetic values, never a real third-party mint, wallet, or handle.
-- The em-dash and en-dash characters are banned in everything you write.
-- Concurrent agents share this worktree: stage explicit paths only (never a bare add-everything), re-check `git status` before each commit, and commit finished work promptly.
-- Before committing, run `npm run check:rules -- --paths <files you touched>`. It must exit 0.
-
-## Why this matters
-
-`/mcp-tools` lists our servers but gives a visitor no way to tell a polished, published, tested server from a stub. The owner's `lyra-registry` scored tools with a weighted trust rubric and A, B, F grades. The rubric ports cleanly to facts we can check mechanically.
-
-## Step 0: re-derive the current state
-
- sed -n 1,60p scripts/build-mcp-catalog.mjs
- node -e "const c=require('./public/mcp-catalog.json');console.log(Object.keys(c), (c.servers||c).length)"
- sed -n 1,40p scripts/lib/mcp-safety-check.mjs
-
-## Tasks
-
-1. A grade per server from checkable facts: published on npm, listed in the registry, tools declared, hosted remote reachable, README and LICENSE present, prompts and resources declared, safety annotations verified by `mcp-safety-check`, Glama claim present. Each fact contributes a documented weight.
-2. Store the grade and its breakdown in `public/mcp-catalog.json`.
-3. Show the grade with a tooltip breakdown on `/mcp-tools`; allow sorting by grade.
-4. Docs: the rubric in `docs/mcp.md`.
-
-## Definition of done
-
-- [ ] `npm run build:mcp-catalog` and `npm run audit:mcp-catalog` pass; a test asserts every server has a grade and breakdown.
-- [ ] `/mcp-tools` browser-verified with sorting and tooltips; docs and changelog.
-
-## Never blocked
-
-| Blocker | Resolution (act, do not ask) |
-|---|---|
-| A fact needs network at build time | Compute network facts in the sentinel (order 529) and read its stored results; build stays offline-safe. |
-
-## Close out (required)
-
-1. Verify every Definition of done line with the actual command output in front of you. Never claim a line you did not verify.
-2. Commit with explicit paths and a subject that describes the diff (house style: `type(scope): what changed and why a reader cares`). User-visible work also gets a `data/changelog.json` entry and `npm run build:pages`; a new surface gets its `STRUCTURE.md` row and its `data/pages.json` entry.
-3. If every Definition of done line passes, delete this file in that same commit (`git rm prompts/finish/531-integrate-32-mcp-catalog-grades.md`) and append a dated entry to [_context/integrate-PROGRESS.md](_context/integrate-PROGRESS.md) with the commit SHAs. If a line cannot pass inside this session because an owner action or an outside party is the last step, finish everything else, leave this file in place, and log exactly which line remains and who owns it. Never delete this file on a partial.
-4. Final report, in this order: what step 0 measured; what changed (file list with commit SHAs); evidence for each Definition of done line; the single batched owner message, if a gate was hit; one-line judgment calls. No trailing questions.
diff --git a/prompts/finish/_context/integrate-PROGRESS.md b/prompts/finish/_context/integrate-PROGRESS.md
index e42cd5606..799574af6 100644
--- a/prompts/finish/_context/integrate-PROGRESS.md
+++ b/prompts/finish/_context/integrate-PROGRESS.md
@@ -1365,6 +1365,10 @@ Step 0 measured: `packages/readme-3d` had no product surface; its README forge e
Step 0 measured: both callers read X through oEmbed (text only, no metrics or media). The shared keyless reader `api/_lib/x-syndication.js` (X syndication, then fxtwitter, then guest-token GraphQL when `X_BEARER_TOKEN` is set) now backs `handleSocialXPostImpact` in `api/pump-fun-mcp.js` and `src/social/x-post-impact.js`; `POST /api/x-video/extract` and the host-locked `GET /api/x-video/download` proxy (https `video.twimg.com` and `pbs.twimg.com` only, no credentials, ports or redirects) plus `pages/x-video.html` and `src/x-video.js` ship the free tool. Code landed in `0ad88fb38`, `22a5a1f49`, `ed75c5946` and `7718bf4fb` (all swept in by concurrent agents). Evidence: `tests/x-post-impact.test.js`, `tests/x-syndication.test.js`, `tests/api/x-video.test.js` and `tests/pumpfun-mcp-tools.test.js` 49 of 49 pass, including host refusals; `check:rules` exits 0; a local server on a spare port returned all mp4 variants (1080p, 720p, 480p and lower) for a real public video post, `no_video` for a text post, streamed a real mp4 through the proxy with 200, and answered 400 `invalid_media_url` for example.com. Open: (1) browser verification of the page and `npm run check:pages`, which needs a Vite build; the shared `node_modules` was being rewritten by a concurrent `npm install` (vite missing, then `@upstash/qstash` and `three` unresolved) for the whole session, so `dist/` is empty of every recent page. Owner: whoever next has a stable install; run `npx vite build && npm run check:pages` and open `/x-video`. The order file stays until then. The live check used a public Tripo post because the recent @trythreews posts the timeline returned contain no video.
-## 2026-10-23: order 525 (integrate 26, Anchor IDL to skill), code done, live decode on a cluster open
+## 2026-10-23: order 537 (integrate 38, market anomaly feed), done
-Step 0 measured: `packages/skill-import` (order 524) had shipped; no Borsh coder existed in `api/`, `src/` or `packages/`; `contracts/idl/` holds `skill_license.json` and `agent_invocation.json`. Shipped `generateAnchorBundle`, `importFromAnchorIdl` and `fetchOnchainIdl` in `packages/skill-import/src/anchor-idl.js` plus the self-contained Borsh codec `anchor-runtime.js` appended to the generated `handlers.js` (the skill sandbox allows no imports), `POST /api/skills/import { idl_url | program_id, cluster? }`, `kind: 'anchor'` drafts (migration `20261023120000_openapi_skill_imports_anchor_kind.sql`, not yet applied), docs and changelog. Tools: `read_<account>`, `preview_<ix>` (read-only, never signs) and `send_<ix>` (`destructiveHint: true`, returns `confirmation_required` with the unsigned instruction); the risk panel flags token movement, SOL spend, account closure and authority changes. Code landed in `b00ee311a`, `22a5a1f49`, `e75ee5681` and `d3f6a7ddd` (swept in by concurrent agents). Evidence: `packages/skill-import/tests/anchor-idl.test.js` 14 of 14 and `tests/openapi-skill-import.test.js` pass; generated bundles for both IDLs pass `validateBundle`; decoded Marketplace and SkillLicense accounts and encoded `mint_skill_license` data match `@coral-xyz/anchor`'s `BorshAccountsCoder` and `BorshInstructionCoder` byte for byte; a spy wallet and a throwing `ctx.secrets` prove preview and send never sign. Open: `skill_license` (`EdngSwxmDktyrr4phwGEZnCXEoQ27vgnBtowjhKa7Wr8`) returns a null account on both devnet and mainnet-beta, so the live decode line needs a devnet deploy (`contracts/skill-license/DEPLOYMENT.md`, owner-gated spend); the check ran on account bytes produced by the reference Anchor coder instead. The order file stays until that passes. Owner: deploy to devnet, then run `read_marketplace` with `cluster: 'devnet'`.
+Step 0 measured: no upstream tarball was needed. The detectors (modified z-score with MAD, EWMA deviation, rate of change) were written from the algorithm description in `api/_lib/market-anomaly.js` and fed from real series: CoinGecko hourly sparklines (top 100), DeFiLlama Solana TVL and Solana stablecoin supply history (built first), the stablecoin board, and Hyperliquid funding history plus open interest snapshots (no public OI history, so the cron records `market_anomaly_samples` each tick). Gas and whale processors were not built (gas is EVM-only, whale detection already exists in Oracle). Shipped in `35afe5904` (engine, sources, store, cron, migration, paid route, alert kind end to end, alerts-mcp), `22a5a1f49` and `2b7bb3921` (registry entry, fixtures and tests swept in by concurrent commits), `3f1eb493d` (ring catalog), plus the docs commit that closes this order. Migration `20261023140000_market_anomalies.sql` applied alone with `--apply --file`; 20 unrelated migrations stay pending. Evidence: 32 detector and alert tests pass on live-recorded fixtures; the cron run wrote 10 real events (stablecoin pegs) and a rerun wrote 0 (dedupe), skipping 20 open interest series as `insufficient_history` (have 2, need 29); `/api/x402/market` lists `market-anomalies` and the route answers 402. Open interest begins scoring after about 7 hours of snapshots.
+
+## 2026-10-23: order 531 (integrate 32) MCP catalog trust grades
+
+Every server in `public/mcp-catalog.json` now carries `trust` (0-100 score, A-F grade, per-fact breakdown) from `scripts/lib/mcp-trust.mjs`; `/mcp-tools` shows the grade with a tooltip breakdown and sorts by it; rubric in `docs/mcp.md`. Commits: `2b7bb3921` (rubric, builder, page, test and docs, swept in by a concurrent commit), `e4bdbdf34` (regenerated catalog), plus the commit that adds the changelog entry and closes this order. Evidence: `build-mcp-catalog --check` and `build-mcp-tool-history --check` pass at 704 tools on 51 servers, `tests/mcp-catalog.test.js` 10 passed, browser-verified sort and tooltips with no console errors. Network facts (npm live, registry listing, endpoint reachable) are deliberately left to the registry sentinel; the rubric scores local publish-readiness only.
diff --git a/public/changelog-recent.json b/public/changelog-recent.json
index f43c5f972..a9e18371b 100644
--- a/public/changelog-recent.json
+++ b/public/changelog-recent.json
@@ -1,11 +1,11 @@
{
- "generated_at": "2026-10-23T22:15:38.567Z",
+ "generated_at": "2026-10-23T22:24:41.310Z",
"generated_by": "scripts/build-page-index.mjs from data/pages.json + data/changelog.json",
"site": {
"name": "three.ws",
"url": "https://three.ws"
},
- "total": 4141,
+ "total": 4142,
"complete": false,
"full_feed": "/changelog.json",
"entries": [
@@ -343,6 +343,16 @@
"sdk"
]
},
+ {
+ "date": "2026-10-23",
+ "type": "update",
+ "title": "Trust grades for every MCP server",
+ "summary": "/mcp-tools now grades each of the 51 MCP servers from A to F on a 0-100 score built from facts the build checks: tools declared and described, safety annotations, an installable package or mounted endpoint, a registry manifest, a README, a license, tests, prompts, resources and the Glama claim. Hover or focus a grade for the full breakdown, and click the Trust column to sort. The rubric and weights are documented.",
+ "link": "/docs/mcp",
+ "tags": [
+ "feature"
+ ]
+ },
{
"date": "2026-10-23",
"type": "update",
@@ -1551,16 +1561,6 @@
"feature",
"desktop"
]
- },
- {
- "date": "2026-10-11",
- "type": "update",
- "title": "The homepage is now the Field: one live map of every agent, launch and wallet move",
- "summary": "The front page no longer opens a web desktop. It opens the Field, a zoomable map of the whole platform built from live data. Every public agent is a star in the core, newest at the centre, and turns into its avatar as you zoom in; around it sit ten districts (Create, Forge, Gallery, Launchpad, Trenches, Pulse, Marketplace, Chat, Agent Studio, Console). From afar each shows one live readout; zoom in and it opens into a card with real content: the newest avatars, coins agents launched, the $THREE price, market cap, holders and contract address, the latest agent wallet moves, the skills agents carry most. New wallet moves from Pulse fly across the map as comets and land in their district. Search finds any place, agent or page; opening one loads the real page in place, with its own shareable /?go= link, and Back or Esc returns you to the map. It works with mouse, touch, pinch and keyboard, in light and dark, and lays itself out for phones. The windowed desktop is still at /?view=desktop and the 3D campus at /?world=1.",
- "link": "/",
- "tags": [
- "feature"
- ]
}
]
}
diff --git a/public/changelog.json b/public/changelog.json
index 0a36e326d..8016481a2 100644
--- a/public/changelog.json
+++ b/public/changelog.json
@@ -1,5 +1,5 @@
{
- "generated_at": "2026-10-23T22:15:38.542Z",
+ "generated_at": "2026-10-23T22:24:41.286Z",
"generated_by": "scripts/build-page-index.mjs from data/pages.json + data/changelog.json",
"site": {
"name": "three.ws",
@@ -340,6 +340,16 @@
"sdk"
]
},
+ {
+ "date": "2026-10-23",
+ "type": "update",
+ "title": "Trust grades for every MCP server",
+ "summary": "/mcp-tools now grades each of the 51 MCP servers from A to F on a 0-100 score built from facts the build checks: tools declared and described, safety annotations, an installable package or mounted endpoint, a registry manifest, a README, a license, tests, prompts, resources and the Glama claim. Hover or focus a grade for the full breakdown, and click the Trust column to sort. The rubric and weights are documented.",
+ "link": "/docs/mcp",
+ "tags": [
+ "feature"
+ ]
+ },
{
"date": "2026-10-23",
"type": "update",
diff --git a/public/changelog.xml b/public/changelog.xml
index ad452bd7f..58f771c6f 100644
--- a/public/changelog.xml
+++ b/public/changelog.xml
@@ -270,6 +270,14 @@
<category>feature</category>
<description>The CLI can now pay a 402 itself. `three-ws pay &lt;url&gt;` reads an x402 v2, x402 v1 or MPP Solana challenge, prefers Solana, signs with a wallet kept in the OS keychain (or an encrypted file behind a passphrase), retries once and prints the body. `--dry-run` shows the payee, exact amount and network and signs nothing. A permissions file sets allowed origins, networks, assets and payees, per-call and per-day maximums and a confirm-above threshold, and a refused challenge exits with a stable code (10) and a reason string. `three-ws pay mcp` serves the same under the same permissions as a stdio MCP server with pay_quote and pay_fetch.</description>
</item>
+ <item>
+ <title>Trust grades for every MCP server</title>
+ <link>https://three.ws/docs/mcp</link>
+ <guid isPermaLink="false">2026-10-23:Trust grades for every MCP server</guid>
+ <pubDate>Fri, 23 Oct 2026 12:00:00 GMT</pubDate>
+ <category>feature</category>
+ <description>/mcp-tools now grades each of the 51 MCP servers from A to F on a 0-100 score built from facts the build checks: tools declared and described, safety annotations, an installable package or mounted endpoint, a registry manifest, a README, a license, tests, prompts, resources and the Glama claim. Hover or focus a grade for the full breakdown, and click the Trust column to sort. The rubric and weights are documented.</description>
+ </item>
<item>
<title>Turn a Solana program's Anchor IDL into an agent skill</title>
<link>https://three.ws/skills/import</link>
diff --git a/public/features.json b/public/features.json
index 038042e46..0da36bbf8 100644
--- a/public/features.json
+++ b/public/features.json
@@ -1,5 +1,5 @@
{
- "generated_at": "2026-10-23T22:15:38.526Z",
+ "generated_at": "2026-10-23T22:24:41.271Z",
"generated_by": "scripts/build-page-index.mjs from data/pages.json",
"site": {
"name": "three.ws",