diff --git a/.github/workflows/context7-refresh.yml b/.github/workflows/context7-refresh.yml
new file mode 100644
index 00000000..241415a6
--- /dev/null
+++ b/.github/workflows/context7-refresh.yml
@@ -0,0 +1,40 @@
+name: Context7 refresh
+
+on:
+  push:
+    branches:
+      - master
+    paths:
+      - "README.md"
+      - "context7.json"
+      - "docs/**"
+
+permissions:
+  contents: read
+
+jobs:
+  refresh:
+    runs-on: ubuntu-latest
+    timeout-minutes: 5
+    steps:
+      - name: Trigger a Context7 refresh
+        env:
+          CONTEXT7_API_KEY: ${{ secrets.CONTEXT7_API_KEY }}
+          LIBRARY_NAME: /${{ github.repository }}
+        run: |
+          status=$(curl --silent --show-error --max-time 180 --output response.json --write-out '%{http_code}' \
+            -X POST https://context7.com/api/v1/refresh \
+            -H "Content-Type: application/json" \
+            -H "Authorization: Bearer $CONTEXT7_API_KEY" \
+            -d "{\"libraryName\": \"$LIBRARY_NAME\"}")
+          cat response.json
+          echo
+          if [ "$status" = "200" ]; then
+            exit 0
+          fi
+          if [ "$status" = "400" ] && jq -e '.error == "too-early"' response.json > /dev/null; then
+            echo "::notice::Context7 refreshed $LIBRARY_NAME recently, so it skipped this refresh."
+            exit 0
+          fi
+          echo "::error::Context7 refresh failed with HTTP $status."
+          exit 1
diff --git a/context7.json b/context7.json
new file mode 100644
index 00000000..b55f755f
--- /dev/null
+++ b/context7.json
@@ -0,0 +1,22 @@
+{
+  "$schema": "https://context7.com/schema/context7.json",
+  "url": "https://context7.com/cardmagic/classifier",
+  "public_key": "pk_veEo2smPvGYMwP1MncN1A",
+  "projectTitle": "Classifier (Ruby gem)",
+  "description": "Text classification in Ruby: Naive Bayes, Logistic Regression, LSI (Latent Semantic Indexing), k-Nearest Neighbors, and TF-IDF vectorization, with a native C extension, streaming training, pluggable persistence, and the classifier and keywords command-line tools. Requires Ruby 3.1+.",
+  "folders": [
+    "docs"
+  ],
+  "excludeFiles": [
+    "AGENTS.md",
+    "CLAUDE.md",
+    "CONTRIBUTING.md"
+  ],
+  "rules": [
+    "Install with gem 'classifier' in the Gemfile, or brew install classifier for the command-line tools only. The gem requires Ruby 3.1 or newer.",
+    "Start with Classifier::Bayes for text classification. Use Classifier::LogisticRegression for calibrated probabilities, Classifier::LSI for semantic search and related documents, and Classifier::KNN for nearest-example votes.",
+    "Classifier::TFIDF must be fitted on a corpus before it transforms text; transform raises Classifier::NotFittedError before a fit. The keywords command fits a vocabulary first and reads that model in later commands.",
+    "Save and load trained models with the JSON persistence API. Prefer it to Marshal, and load a marshalled model only from a source you trust.",
+    "For data larger than memory, use train_from_stream, which reads one document per line from an IO stream. Classifier::TFIDF uses fit_from_stream. Use train_batch only for documents already in memory; Classifier::LSI and Classifier::KNN name it add_batch."
+  ]
+}