@tensorfeed/mcp-server 1.38.0 → 2.0.0

package/CHANGELOG.md

@@ -2,6 +2,104 @@
 
 All notable changes to the [TensorFeed.ai MCP server](https://github.com/RipperMercs/tensorfeed-mcp). Free tools work without configuration; premium tools require a bearer token via the `TENSORFEED_TOKEN` env var. Buy credits at [tensorfeed.ai/developers/agent-payments](https://tensorfeed.ai/developers/agent-payments).
 
+## 2.0.0 - 2026-06-07
+
+Major tool-surface rework. BREAKING: the catalog drops from about 78 tools to
+exactly 24. The old surface had grown wide enough that agents picked the wrong
+tool or stalled across near-duplicate names; 24 sharp tools make tool-selection
+reliable. No HTTP endpoint was removed. Everything still works at
+tensorfeed.ai; this release only changes which slices are first-class MCP tools.
+
+### Migration
+
+If your code called any of the names below, switch to the replacement:
+
+- The eight signed verdicts collapse from preview plus paid pairs into one tool
+  each with a `tier` parameter. `<family>_verdict_preview` and
+  `<family>_verdict` are now a single `<family>_verdict`, where `tier` is
+  `preview` (free, the default) or `full` (1 credit, AFTA-signed). Affected
+  families: route, provider_reliability, x402_settlement, x402_publisher,
+  stack_safety, benchmark_trust, failover, ssvc.
+- The four time series collapse from free plus paid pairs into one tool each
+  with a `days` parameter. `pricing_series_free` and `pricing_series` are now a
+  single `pricing_series`; the same applies to `benchmark_series`,
+  `status_uptime`, and `status_leaderboard`. Pass `days` 1 to 7 for the free
+  window or 8 to 90 for the paid window.
+- The four watch creators merge into one `create_watch` with a `type`
+  parameter. `create_price_watch`, `create_status_watch`, `create_digest_watch`,
+  and `create_leaderboard_rank_watch` are now `create_watch` with `type` set to
+  `price`, `status`, `digest`, or `leaderboard_rank`.
+- `get_ai_today` folds into `get_ai_news`. Pass `digest: true` for the daily
+  brief that `get_ai_today` used to return.
+- `get_account_balance` and `get_account_usage` merge into one `account_status`
+  that returns both the balance and the usage view.
+
+### Removed from the tool list (still live as HTTP endpoints)
+
+About 40 long-tail standalone tools were dropped from the MCP surface, including
+the model-deprecation, agent-activity, IOC, AFTA-certification,
+agent-reputation, agent-readiness, crawler-access, MCP-registry-snapshot,
+probe, paper and trending feeds, Reddit, OpenRouter, ecosystem-today,
+earthquake, weather-alert, agent-opportunity, AI-CVE, company-filings, x402
+market, and the standalone routing, news-search, cost-projection, and
+agents-directory tools. None of these went away on the server. Each one is
+still callable as an HTTP endpoint on tensorfeed.ai, and the new
+`find_tensorfeed_data` tool helps an agent discover the right one on demand.
+
+### Added
+
+- `find_tensorfeed_data`: a discovery tool. Describe the AI-ecosystem data you
+  want and it returns the matching TensorFeed endpoints, with the path and
+  parameters to call, so an agent can reach the full catalog without a dedicated
+  tool for every feed.
+
+### The 24 tools
+
+`get_ai_news`, `get_ai_status`, `is_service_down`, `get_model_pricing`,
+`route_verdict`, `provider_reliability_verdict`, `x402_settlement_verdict`,
+`x402_publisher_verdict`, `stack_safety_verdict`, `benchmark_trust_verdict`,
+`failover_verdict`, `ssvc_verdict`, `whats_new`, `compare_models`,
+`provider_deepdive`, `pricing_series`, `benchmark_series`, `status_uptime`,
+`status_leaderboard`, `account_status`, `list_watches`, `create_watch`,
+`delete_watch`, `find_tensorfeed_data`.
+
+## 1.39.0 - 2026-06-04
+
+Surfaces TensorFeed's full signed-verdict family in the stdio server. Adds the
+seven verdicts below, each shipping as a free preview tool plus a 1-credit
+($0.02) signed premium tool with an AFTA-signed receipt. This brings the catalog
+to 79 tools (48 free, 31 premium):
+- `provider_reliability_verdict_preview` / `provider_reliability_verdict`:
+  the signed dependability ruling over TensorFeed's own latency and availability
+  probes, scoring availability and tail consistency to name the safest provider
+  to build on.
+- `x402_settlement_verdict_preview` / `x402_settlement_verdict`: the settlement
+  momentum, concentration, and leading-publisher read over the on-chain x402
+  index for a 24h, 7d, or 30d window.
+- `x402_publisher_verdict_preview` / `x402_publisher_verdict`: the signed trust
+  verdict on one publisher domain (actively settling, recently quiet, registered
+  with no settlement, unreachable, no Base payTo, or not indexed) before you
+  pay it.
+- `stack_safety_verdict_preview` / `stack_safety_verdict`: the GO, HOLD, or
+  BLOCK deploy gate over a list of package@version pins, with the matched-CVE
+  and KEV evidence for the worst offender.
+- `benchmark_trust_verdict_preview` / `benchmark_trust_verdict`: the trust band
+  and 0 to 100 score for a benchmark or category, flagging saturation,
+  contamination, and held-out status.
+- `failover_verdict_preview` / `failover_verdict`: when a provider is degraded,
+  the signed ruling on the single best operational provider to fail over to,
+  with ranked alternatives.
+- `ssvc_verdict_preview` / `ssvc_verdict`: the CISA SSVC Act, Attend, Track, or
+  Track* decision for one CVE, with the decision points and a live KEV
+  cross-check.
+
+Adds contextual pointers so an agent discovers the matching verdict from a tool
+it already uses: `is_service_down` points to `failover_verdict`,
+`get_ai_cves_latest` points to `ssvc_verdict` and `stack_safety_verdict`,
+`get_x402_publishers` points to `x402_publisher_verdict`, and
+`status_leaderboard_free` points to `provider_reliability_verdict`. No behavior
+change to those four tools.
+
 ## 1.38.0 - 2026-06-03
 
 Description-only release. Sharpens the six foundational free tools (get_ai_news,
@@ -290,46 +388,46 @@ All three are backed by the worker endpoints shipped in the same monorepo commit
 ## 1.11.0 - 2026-04-29
 
 ### Added
-- `probe_latest` tool — last 24 hours of measured LLM endpoint latency and availability per provider (Anthropic, OpenAI, Google, Mistral, Cohere). TensorFeed pings each provider's chat completion endpoint every 15 min and records ttfb / total / status. The data is unique because it is measured, not self-reported. Free, no auth.
-- `probe_series` tool — daily SLA series for one provider with ttfb/total p50/p95/p99 and incident-hour count, 90-day max range. 1 credit per call. Pairs with `premium_routing` for picking a model whose SLA you can verify.
+- `probe_latest` tool: last 24 hours of measured LLM endpoint latency and availability per provider (Anthropic, OpenAI, Google, Mistral, Cohere). TensorFeed pings each provider's chat completion endpoint every 15 min and records ttfb / total / status. The data is unique because it is measured, not self-reported. Free, no auth.
+- `probe_series` tool: daily SLA series for one provider with ttfb/total p50/p95/p99 and incident-hour count, 90-day max range. 1 credit per call. Pairs with `premium_routing` for picking a model whose SLA you can verify.
 
 ## 1.10.0 - 2026-04-29
 
 ### Added
-- `mcp_registry_snapshot` tool — today's summary of the official Model Context Protocol server registry: total servers, by-status breakdown, top namespaces, 1-day deltas (newly added, reactivated, deprecated). Free, no auth.
-- `mcp_registry_series` tool — multi-day time series of MCP registry growth and churn. Range capped at 90 days, default 30 days back. 1 credit per call.
+- `mcp_registry_snapshot` tool: today's summary of the official Model Context Protocol server registry: total servers, by-status breakdown, top namespaces, 1-day deltas (newly added, reactivated, deprecated). Free, no auth.
+- `mcp_registry_series` tool: multi-day time series of MCP registry growth and churn. Range capped at 90 days, default 30 days back. 1 credit per call.
 
 The new tools are backed by a fresh daily cron capture of registry.modelcontextprotocol.io. The registry itself is open data, but a 30/90-day trend requires daily capture started weeks before the question is asked.
 
 ## 1.7.0 - 2026-04-27
 
 ### Added
-- `compare_models` tool — side-by-side comparison of 2-5 AI models with normalized benchmarks (union-of-keys with null for missing scores) and rankings (cheapest blended, most context, per-benchmark leaderboard). 1 credit per call.
+- `compare_models` tool: side-by-side comparison of 2-5 AI models with normalized benchmarks (union-of-keys with null for missing scores) and rankings (cheapest blended, most context, per-benchmark leaderboard). 1 credit per call.
 
 ## 1.6.0 - 2026-04-27
 
 ### Added
-- `provider_deepdive` tool — one provider's full profile in a single call: live status with components, all models with pricing + tier + benchmark scores joined, recent news, agent traffic. 1 credit per call.
+- `provider_deepdive` tool: one provider's full profile in a single call: live status with components, all models with pricing + tier + benchmark scores joined, recent news, agent traffic. 1 credit per call.
 
 ## 1.5.0 - 2026-04-27
 
 ### Added
-- `create_digest_watch` tool — register a scheduled daily/weekly digest watch. Fires HMAC-signed POST to your callback URL with a curated pricing-changes summary regardless of whether anything dramatic happened. 1 credit per registration.
+- `create_digest_watch` tool: register a scheduled daily/weekly digest watch. Fires HMAC-signed POST to your callback URL with a curated pricing-changes summary regardless of whether anything dramatic happened. 1 credit per registration.
 
 ## 1.4.0 - 2026-04-27
 
 ### Added
-- `forecast` tool — conservative linear-regression forecast for a price field or benchmark score with 95% prediction interval and confidence label. 1 credit per call.
+- `forecast` tool: conservative linear-regression forecast for a price field or benchmark score with 95% prediction interval and confidence label. 1 credit per call.
 
 ## 1.3.0 - 2026-04-27
 
 ### Added
-- `cost_projection` tool — project workload cost across 1-10 AI models with four time horizons and cheapest-monthly ranking. 1 credit per call.
+- `cost_projection` tool: project workload cost across 1-10 AI models with four time horizons and cheapest-monthly ranking. 1 credit per call.
 
 ## 1.2.0 - 2026-04-27
 
 ### Added
-- `news_search` tool — full-text search over the AI news corpus with date/provider/category filters and relevance scoring. 1 credit per call.
+- `news_search` tool: full-text search over the AI news corpus with date/provider/category filters and relevance scoring. 1 credit per call.
 
 ## 1.1.0 - 2026-04-27
 
@@ -339,7 +437,7 @@ The new tools are backed by a fresh daily cron capture of registry.modelcontextp
 - `mcp-server/server.json` manifest authored against the official MCP registry schema (`ai.tensorfeed/mcp-server`).
 - README expanded with per-tier tool tables and `env` config examples for Claude Desktop / Code.
 
-## 1.0.0 — 2026-04-26 (initial release)
+## 1.0.0 - 2026-04-26 (initial release)
 
 ### Added
 - 5 free tools: `get_ai_news`, `get_ai_status`, `is_service_down`, `get_model_pricing`, `get_ai_today`. No auth, no install beyond `npx -y @tensorfeed/mcp-server`.

package/README.md

@@ -43,22 +43,38 @@ Swap `task` for `reasoning`, `creative`, or `general`, or pass `?model=<id-or-na
 
 ### The agent path (MCP)
 
-With `@tensorfeed/mcp-server` installed, an agent gets two tools. Call the free one for the pick, then the paid one when it needs to defend the choice:
+With `@tensorfeed/mcp-server` installed, an agent gets one `route_verdict` tool with a `tier` parameter. Call it with the default free tier for the pick, then `tier="full"` when it needs to defend the choice:
 
 ```text
-# Free taste: the top pick + reasoning, no token (10/IP/day)
-route_verdict_preview({ task: "code" })
+# Free taste: the top pick + reasoning, no token (10/IP/day). tier defaults to "preview".
+route_verdict({ task: "code" })
 
 # 1 credit: ranked runners-up, constraint filters, AFTA-signed receipt
-route_verdict({ task: "code", max_latency_p95_ms: 1500, budget: 8, min_quality: 0.6 })
+route_verdict({ task: "code", tier: "full", max_latency_p95_ms: 1500, budget: 8, min_quality: 0.6 })
 ```
 
-`route_verdict` adds the ranked runners-up, the constraint filters (`max_latency_p95_ms`, `budget`, `min_quality`, `require_operational`, `exclude_deprecated`), and the AFTA-signed receipt the agent can audit later. Credits come from [tensorfeed.ai/developers/agent-payments](https://tensorfeed.ai/developers/agent-payments).
+`tier="full"` adds the ranked runners-up, the constraint filters (`max_latency_p95_ms`, `budget`, `min_quality`, `require_operational`, `exclude_deprecated`), and the AFTA-signed receipt the agent can audit later. Credits come from [tensorfeed.ai/developers/agent-payments](https://tensorfeed.ai/developers/agent-payments).
 
 ### Why it matters
 
 Models, prices, and latency move week to week. `route_verdict` is one signed call an agent can act on now and later prove why it routed the way it did, without rebuilding the comparison from scratch each time.
 
+## Catalog
+
+24 tools. The core flagships, the eight signed verdicts, the time-series tools, and the webhook watches are dedicated tools; the rest of the 100+ TensorFeed endpoints are reachable through the `find_tensorfeed_data` discovery tool and callable over HTTP. Free tiers need no token; paid tiers charge USDC on Base via x402 and return an AFTA-signed receipt. The full tool reference lives in the [standalone repo](https://github.com/RipperMercs/tensorfeed-mcp).
+
+### Verdict family
+
+Eight signed decisions (route_verdict is featured above). Each is a single tool with a `tier` parameter: `tier="preview"` (default) is free, `tier="full"` costs 1 credit ($0.02) and adds the full ranking and an AFTA-signed receipt:
+
+- **provider_reliability_verdict**: the safest AI provider to build on, ranked by availability and tail consistency over TensorFeed's own probes.
+- **x402_settlement_verdict**: the x402 settlement momentum, concentration, and leading publisher over a 24h, 7d, or 30d window.
+- **x402_publisher_verdict**: a signed trust verdict on one publisher domain before you pay it.
+- **stack_safety_verdict**: a GO, HOLD, or BLOCK deploy gate over your package@version pins, with the worst CVE or KEV match.
+- **benchmark_trust_verdict**: a trust band and 0 to 100 score for a benchmark, flagging saturation, contamination, and held-out status.
+- **failover_verdict**: when a provider is degraded, the single best operational provider to fail over to, with ranked alternatives.
+- **ssvc_verdict**: the CISA SSVC Act, Attend, Track, or Track* decision for one CVE, with a live KEV cross-check.
+
 ## Publish a new version
 
 From the main tensorfeed repo:

package/dist/discovery.d.ts

@@ -0,0 +1,9 @@
+export interface EndpointRow {
+    key: string;
+    path: string;
+    description: string;
+}
+/** Flatten the /api/meta `api` object ({key: 'path (desc)'}) into rows. */
+export declare function flattenCatalog(meta: unknown): EndpointRow[];
+/** Keyword-overlap score of a query against each row's key + description. Returns top N with score > 0. */
+export declare function scoreEndpoints(rows: EndpointRow[], query: string, limit: number): EndpointRow[];

package/dist/discovery.js

@@ -0,0 +1,38 @@
+/** Flatten the /api/meta `api` object ({key: 'path (desc)'}) into rows. */
+export function flattenCatalog(meta) {
+    const api = meta?.api;
+    if (!api || typeof api !== 'object')
+        return [];
+    const rows = [];
+    for (const [key, val] of Object.entries(api)) {
+        if (typeof val !== 'string')
+            continue;
+        const m = val.match(/^(\S+)\s*(.*)$/);
+        const path = m ? m[1] : val;
+        const description = m ? m[2].replace(/^\(/, '').replace(/\)$/, '') : '';
+        rows.push({ key, path, description });
+    }
+    return rows;
+}
+const STOP = new Set(['the', 'a', 'an', 'for', 'of', 'to', 'and', 'or', 'in', 'on', 'with', 'me', 'get', 'find', 'data', 'i', 'need', 'want', 'my']);
+function tokenize(s) {
+    return s.toLowerCase().split(/[^a-z0-9]+/).filter((t) => t.length > 2 && !STOP.has(t));
+}
+/** Keyword-overlap score of a query against each row's key + description. Returns top N with score > 0. */
+export function scoreEndpoints(rows, query, limit) {
+    const q = new Set(tokenize(query));
+    if (q.size === 0)
+        return [];
+    const scored = rows
+        .map((r) => {
+        const hay = tokenize(r.key + ' ' + r.description);
+        let score = 0;
+        for (const t of hay)
+            if (q.has(t))
+                score += 1;
+        return { r, score };
+    })
+        .filter((x) => x.score > 0)
+        .sort((a, b) => b.score - a.score);
+    return scored.slice(0, limit).map((x) => x.r);
+}