diff --git a/.github/scripts/dep_table.py b/.github/scripts/dep_table.py
new file mode 100644
index 000000000..0073be3ca
--- /dev/null
+++ b/.github/scripts/dep_table.py
@@ -0,0 +1,131 @@
+#!/usr/bin/env python3
+"""Print a markdown table of Boost dependency changes between two CI log archives.
+
+Usage: dep_table.py BASELINE.zip AFTER.zip
+
+Each archive is a GitHub Actions run-log zip. The `deps` job of the
+dependency-report workflow prints two marker-delimited boostdep reports into
+its log:
+
+  ===DEP-BRIEF-START===    boostdep --brief graph      ===DEP-BRIEF-END===
+  ===DEP-PRIMARY-START===  boostdep graph (primary)    ===DEP-PRIMARY-END===
+
+From those it derives two metrics and prints their delta vs the baseline:
+  - the set of transitive Boost modules graph depends on (from --brief:
+    Primary + Secondary sections = the full transitive closure), and
+  - the header-inclusion weight of each direct dependency (from the primary
+    report: how many distinct boost/graph files pull that dependency in).
+Weights are the live signal, they drop toward zero as coupling is removed;
+the module count is the coarse target that only moves when a dep hits zero.
+"""
+import re
+import sys
+import zipfile
+
+BRIEF = ("===DEP-BRIEF-START===", "===DEP-BRIEF-END===")
+PRIMARY = ("===DEP-PRIMARY-START===", "===DEP-PRIMARY-END===")
+# Module names as boostdep prints them, e.g. "numeric~conversion".
+MODULE = re.compile(r"[A-Za-z0-9_.~-]+")
+TIMESTAMP = re.compile(r"^\d{4}-\d\d-\d\dT[\d:.]+Z\s?")  # GitHub log line prefix
+ANSI = re.compile(r"\x1b\[[0-9;]*m")
+
+
+def read_clean_lines(zip_path):
+    """Top-level per-job logs, with GitHub timestamp and ANSI prefixes stripped."""
+    lines = []
+    with zipfile.ZipFile(zip_path) as z:
+        for entry in z.namelist():
+            if "/" in entry or not entry.endswith(".txt"):
+                continue
+            for ln in z.read(entry).decode("utf-8", "replace").splitlines():
+                lines.append(TIMESTAMP.sub("", ANSI.sub("", ln)))
+    return lines
+
+
+def block(lines, markers):
+    """Lines strictly between the marker lines, matched exactly.
+
+    Exact matching is what skips the echoed `echo "<marker>"` command lines
+    GitHub prepends to a run step, so we capture boostdep's real stdout.
+    """
+    start, end = markers
+    out, capturing = [], False
+    for ln in lines:
+        s = ln.strip()
+        if s == start:
+            capturing = True
+            continue
+        if s == end and capturing:
+            break
+        if capturing:
+            out.append(ln)
+    return out
+
+
+def parse_brief(lines):
+    """--brief output -> set of module names (one bare name per line)."""
+    mods = set()
+    for ln in lines:
+        s = ln.strip()
+        if not s or s.startswith("#") or s.lower().startswith("brief dependency"):
+            continue
+        if MODULE.fullmatch(s):
+            mods.add(s)
+    return mods
+
+
+def parse_weights(lines):
+    """Primary report -> {dependency: number of distinct graph files that pull it in}."""
+    weights, cur = {}, None
+    for ln in lines:
+        head = re.match(r"^([A-Za-z0-9_.~-]+):\s*$", ln)  # "module:" at column 0
+        if head:
+            cur = head.group(1)
+            weights.setdefault(cur, set())
+            continue
+        frm = re.match(r"^\s+from\s+(.+?)\s*$", ln)
+        if frm and cur is not None:
+            weights[cur].add(frm.group(1))
+    return {k: len(v) for k, v in weights.items()}
+
+
+def main(base_zip, pr_zip):
+    bl, pl = read_clean_lines(base_zip), read_clean_lines(pr_zip)
+    base_mods = parse_brief(block(bl, BRIEF))
+    pr_mods = parse_brief(block(pl, BRIEF))
+    base_w = parse_weights(block(bl, PRIMARY))
+    pr_w = parse_weights(block(pl, PRIMARY))
+
+    # Header-inclusion weights: the live signal. Only rows that changed, most-reduced first.
+    print("**Header-inclusion weights** (graph files pulling each direct dependency in):")
+    print()
+    changed = []
+    for dep in base_w.keys() | pr_w.keys():
+        b, p = base_w.get(dep, 0), pr_w.get(dep, 0)
+        if b != p:
+            changed.append((p - b, dep, b, p))
+    if changed:
+        print("| Dependency | develop | PR | Δ |")
+        print("|-----|--------:|---:|----:|")
+        for d, dep, b, p in sorted(changed):  # reductions (negative delta) first
+            print(f"| {dep} | {b} | {p} | {d:+d} |")
+    else:
+        print("_No header-inclusion-weight changes._")
+    print()
+
+    # Transitive module set: the coarse target.
+    added = sorted(pr_mods - base_mods)
+    removed = sorted(base_mods - pr_mods)
+    nb, np_ = len(base_mods), len(pr_mods)
+    diff = np_ - nb
+    print(f"**Transitive Boost modules:** {nb} → {np_} ({f'{diff:+d}' if diff else '0'})")
+    if added:
+        print(f"- added: {', '.join(added)}")
+    if removed:
+        print(f"- removed: {', '.join(removed)}")
+
+
+if __name__ == "__main__":
+    if len(sys.argv) != 3:
+        sys.exit("usage: dep_table.py BASELINE.zip AFTER.zip")
+    main(sys.argv[1], sys.argv[2])
diff --git a/.github/workflows/dependency-count-comment.yml b/.github/workflows/dependency-count-comment.yml
new file mode 100644
index 000000000..ddf49b0a1
--- /dev/null
+++ b/.github/workflows/dependency-count-comment.yml
@@ -0,0 +1,61 @@
+name: dependency-count-comment
+
+on:
+  workflow_run:
+    workflows: ["dependency-report"]
+    types: [completed]
+
+permissions:
+  actions: read
+  pull-requests: write
+  contents: read
+
+jobs:
+  comment:
+    if: github.event.workflow_run.event == 'pull_request'
+    runs-on: ubuntu-latest
+    env:
+      GH_TOKEN: ${{ secrets.GITHUB_TOKEN }}
+      REPO: ${{ github.repository }}
+      PR_RUN: ${{ github.event.workflow_run.id }}
+      HEAD_SHA: ${{ github.event.workflow_run.head_sha }}
+    steps:
+      - uses: actions/checkout@v4
+
+      - name: Download this PR run's log archive
+        run: gh api "repos/$REPO/actions/runs/$PR_RUN/logs" > pr-logs.zip
+
+      - name: Download latest successful develop run's log archive
+        run: |
+          DEV=$(gh run list --workflow dependency-report --branch develop --status success \
+                  --limit 1 --json databaseId --jq '.[0].databaseId')
+          if [ -z "$DEV" ]; then echo "no develop baseline run found" >&2; exit 1; fi
+          echo "DEV_RUN=$DEV" >> "$GITHUB_ENV"
+          gh api "repos/$REPO/actions/runs/$DEV/logs" > base-logs.zip
+
+      - name: Generate table
+        run: |
+          {
+            echo "Boost dependency footprint vs \`develop\` (auto-generated)."
+            echo "PR run \`$PR_RUN\` vs develop run \`$DEV_RUN\` (\`${HEAD_SHA:0:10}\`)."
+            echo
+            python3 .github/scripts/dep_table.py base-logs.zip pr-logs.zip
+          } > table.md
+
+      - name: Resolve PR number
+        id: pr
+        run: |
+          num=$(gh api "repos/$REPO/pulls?state=open&per_page=100" \
+                  --jq "map(select(.head.sha==\"$HEAD_SHA\"))[0].number")
+          if [ -z "$num" ]; then
+            echo "no open PR found with head $HEAD_SHA" >&2
+            exit 1
+          fi
+          echo "num=$num" >> "$GITHUB_OUTPUT"
+
+      - name: Post or update sticky comment
+        uses: marocchino/sticky-pull-request-comment@v2
+        with:
+          header: dependency-counts
+          number: ${{ steps.pr.outputs.num }}
+          path: table.md
diff --git a/.github/workflows/dependency-report.yml b/.github/workflows/dependency-report.yml
new file mode 100644
index 000000000..c788d5e06
--- /dev/null
+++ b/.github/workflows/dependency-report.yml
@@ -0,0 +1,42 @@
+name: dependency-report
+
+# Computes graph's Boost dependency footprint with boostdep and prints a
+# marker-delimited report into the job log, on every push and PR. The
+# dependency-count-comment workflow reads these logs (this PR's run and the
+# latest develop run) and posts the diff as a PR comment. Kept separate from
+# CI so ci.yml is untouched.
+
+on: [ push, pull_request ]
+
+jobs:
+  deps:
+    runs-on: ubuntu-24.04
+    steps:
+      - uses: actions/checkout@v4
+      - name: Set up Boost tree and build boostdep
+        run: |
+          git clone -b develop --depth 1 https://github.com/boostorg/boost.git ../boost-root
+          cd ../boost-root
+          cp -r $GITHUB_WORKSPACE/* libs/graph
+          git submodule update --init tools/boostdep
+          # depinst pulls graph's deps; also ensure boostdep's own dep
+          # (Boost.Filesystem) is present since boostdep links against it.
+          python tools/boostdep/depinst/depinst.py --git_args "--jobs 3" graph
+          python tools/boostdep/depinst/depinst.py --git_args "--jobs 3" filesystem
+          ./bootstrap.sh
+          ./b2 headers
+          # boostdep #includes <boost/filesystem.hpp> and links the compiled
+          # library, so build it with b2 (a bare c++ compile can't resolve that).
+          ./b2 tools/boostdep/build
+      - name: Boostdep dependency report
+        working-directory: ../boost-root
+        run: |
+          BOOSTDEP=$(find "$PWD" -type f -name boostdep -perm -u+x 2>/dev/null | grep -v '/src/' | head -1)
+          if [ -z "$BOOSTDEP" ]; then echo "boostdep executable not found after build" >&2; exit 1; fi
+          echo "using boostdep: $BOOSTDEP"
+          echo "===DEP-BRIEF-START==="
+          "$BOOSTDEP" --boost-root . --track-sources --no-track-tests --brief graph
+          echo "===DEP-BRIEF-END==="
+          echo "===DEP-PRIMARY-START==="
+          "$BOOSTDEP" --boost-root . --track-sources --no-track-tests graph
+          echo "===DEP-PRIMARY-END==="