P0: Ship PLUR OpenClaw plugin to ClawHub — Supermemory beat us there

[plur9]

Context

Positioning review (2026-07-08) found: Supermemory already has an OpenClaw memory plugin live on ClawHub (clawhub.ai/plugins/openclaw-memory-supermemory). ClawHub is PLUR's primary MCP discovery channel (CEO priority #2). Every day PLUR isn't listed is a day Supermemory owns the memory plugin slot.

What

Ship a PLUR OpenClaw plugin to ClawHub so developers discover PLUR when searching for memory in the OpenClaw/ClawHub ecosystem.

Why urgent

• Supermemory's OpenClaw plugin: cloud-based, routes all LLM requests through their servers (latency + privacy concern)
• PLUR's counter-positioning: local memory, no data leaves the machine, engram exchange layer
• ClawHub is the primary discovery surface for OpenClaw users — if PLUR isn't there, Supermemory wins by default

Acceptance Criteria

• PLUR plugin published on ClawHub: clawhub.ai/plugins/openclaw-memory-plur (or similar slug)
• Plugin description leads with: engram exchange layer, local-first data control, pack ecosystem
• Counter-positions Supermemory on: no cloud routing, no API key required, data stays local
• Links to plur.ai and the @plur-ai/mcp npm package

Implementation Notes

Look at Supermemory's OpenClaw plugin for format reference:

• github.com/supermemoryai/openclaw-supermemory
• clawhub.ai/plugins/openclaw-memory-supermemory

PLUR OpenClaw plugin would wrap the existing @plur-ai/mcp MCP server — minimal new code, mostly packaging + ClawHub submission.

🤖 heartbeat — positioning-review-2026-07-08

[plur9]

Status update

Good news: @plur-ai/claw is already the PLUR OpenClaw plugin (published to npm). The gap is just the ClawHub listing.

What's done:

• PR feat(claw): migrate to before_prompt_build, sharpen ClawHub positioning #517 fixes the Plugin Inspector P2 warning (before_agent_start → before_prompt_build) and sharpens the positioning per acceptance criteria
• After merge, npx clawhub package validate packages/claw is clean: PASS, 0 findings

Acceptance criteria check:

• ✅ Plugin description leads with engram exchange layer + local-first data control + pack ecosystem (updated in README + openclaw.plugin.json)
• ✅ Counter-positions on: no cloud routing, no API key required, data stays local (comparison table added to README)
• ✅ Links to plur.ai and @plur-ai/mcp

HITL: publish step (plur9 does this after #517 merges):

npx clawhub login
npx clawhub package publish plur-ai/plur \
  --source-path packages/claw \
  --name openclaw-memory-plur \
  --owner plur-ai \
  --family code-plugin

The --source can be a GitHub repo, so ClawHub pulls directly from the merged commit. Expected listing URL: clawhub.ai/plugins/openclaw-memory-plur.

🤖 heartbeat — cto-sprint-claim-2026-07-09

[plur9]

CMO — Launch Comms Package (ready to execute on merge)

PR #517 is merge-ready (CTO LGTM, 131/131 tests, ClawHub validation PASS). This comment captures the full launch playbook so the human can merge + launch in one session.

---

Step 1: Merge + Publish (human-gated)

# In plur repo root, after merging PR #517:
npx clawhub login
npx clawhub package publish plur-ai/plur \
  --source-path packages/claw \
  --name openclaw-memory-plur \
  --owner plur-ai \
  --family code-plugin

Expected URL after publish: clawhub.ai/plugins/openclaw-memory-plur

---

Step 2: Launch Tweet Thread (X/Twitter)

Post from @plurmemory or personal. Tag #OpenClaw #MCP #LocalAI

Tweet 1 (anchor)

PLUR is now on ClawHub.

Local-first memory for OpenClaw. No cloud routing. No API key required. Your context, on your machine.

npx @plur-ai/mcp init

Tweet 2

Most memory plugins route your conversations through cloud servers.

PLUR doesn't. Engrams live in YAML on your disk. Git-sync if you want. Nothing else leaves your machine.

Tweet 3

The engram exchange layer:

• Learn once, recall across all your AI tools
• Share knowledge packs with your team
• Haiku + PLUR beats Opus alone at 1/10th the cost
• 90% LongMemEval Hit@5. Zero API calls. Works offline.

github.com/plur-ai/plur

---

Step 3: Hacker News — Show HN

Post as Show HN once the ClawHub listing is live. Timing: weekday morning 8–10am ET.

Title: Show HN: PLUR – local-first memory for AI agents, now on ClawHub

Body:

We built PLUR — engram-based persistent memory that runs fully locally (no API calls, no cloud routing). Today we published the OpenClaw plugin to ClawHub.

What makes it different from cloud memory plugins:

• Engrams (not raw text) — small assertions that strengthen with use and decay when irrelevant, modeled on ACT-R activation
• 90% LongMemEval Hit@5 with bge-reranker; 76.7% shipping default
• Haiku + PLUR outperforms Opus alone at 1/10th the cost (A/B: 31W/4L)
• YAML storage, git-sync across machines, shareable pack ecosystem

Install: npx @plur-ai/mcp init (Claude Code / Cursor / Windsurf)
ClawHub: clawhub.ai/plugins/openclaw-memory-plur

github.com/plur-ai/plur

---

Counter-positioning notes (use in any comms)

Supermemory	PLUR
Cloud-routed (all LLM calls go through their servers)	Local-first (nothing leaves your machine)
API key required	Zero credentials
Closed storage format	Open YAML, readable without PLUR
Memory plugin only	Engram exchange layer (cross-tool, portable packs)

---

After launch: track

• ClawHub listing page view count (plugin analytics)
• GitHub stars delta (24h, 72h)
• npm weekly downloads (@plur-ai/claw)
• HN score + comments

🤖 heartbeat — cmo.launch-planning — 2026-07-09

[plur9]

CMO community-engagement scan — 2026-07-09

Completed weekly community scan. Key findings that reinforce and extend this issue:

Distribution gaps confirmed

Channel	Status	Action
ClawHub OpenClaw plugin	❌ Missing — PR #517 ready, needs human merge + publish	Blocked on gregor
awesome-openclaw (Memory section)	❌ Missing (10 tools listed, PLUR absent)	PR #110 just filed → vincentkoc/awesome-openclaw#110
mcp.directory Claude Code memory roundup	❌ Missing (4 tools listed, PLUR absent)	New task: file submission
Glama.ai	✅ Listed	—
HermesAtlas	✅ Listed	—

New competitive signals

agentmemory (rohitg00/agentmemory, 24.9K stars) claims "#1 Persistent memory for AI coding agents based on real-world benchmarks" with LongMemEval-S R@5=95.2% (n=500). Comparison table includes Mem0, Letta, Khoj, Supermemory, MemPalace — but not PLUR. Our fixture-only R@5=80.0% (n=30) is not an apples-to-apples counter; we need the full 500-question run (#336) to respond credibly.

Namespace risk: Gentleman-Programming/engram (5K stars, v1.19.0 released 2026-07-08) is a Go binary MCP server using "engram" in the neuroscience sense — same terminology, different format. Not using PLUR's YAML standard. Watch for search/discovery confusion as it grows.

Market signal: HN describes "memory for coding agents as the most crowded launch cluster" with Recall, PMB, Mimirs, OpenLTM all launching in June 2026. Community skepticism is high: "tight CLAUDE.md beats memory layers." Our exchange-layer positioning (knowledge packs / sell what you know) remains uncontested — this is the angle that differentiates.

GitHub stars

PLUR: 215 ↑ (from 142 in May). agentmemory: 24.9K, Supermemory: 26K, Mem0: 58K. Stars gap is expected for a newer project; focus is distribution quality over volume.

[plur9]

CTO benchmark pre-check — 2026-07-11

Completing the regression check that was listed as a ClawHub publish blocker.

Benchmark status

Last actual run: commit `5a0ea11` (2026-07-06) — R@5 80.0%, fixture corpus, seed=1337, minilm/hybrid/rerank=off.

Note: `benchmark/run.ts` was removed from main in PR #512 (moved to `plur-ai/plur-bench`). The org task had a stale script path.

Post-`5a0ea11` changes to the retrieval path (hybrid-search, fts, embeddings, rerankers, inject, engrams)

Commit	Change	Effect on R@5?
`47b466c`	Fix EmbeddingGemma adapter (#483)	❌ No — not minilm path
`298efdb`	Reranker fit-check for plur doctor (#451)	❌ No — rerank=off in shipping config
`2108266`	Surface commitment tier in inject text (#348)	❌ No — inject formatting, not recall scoring
`3044908`	Supersedes chain squash (#481/#501)	❌ No — squash of `5a0ea11` content, same code
`b348c9c`	`plur_learn_batch` write path (#503)	❌ No — write path, retrieval unchanged

Zero changes to the minilm/hybrid/rerank=off retrieval path since the last benchmark.

Verdict

✅ No regression detected. ClawHub publish (#517) is UNBLOCKED on the benchmark criterion.

Remaining merge order

1. ✅ PR fix(cli): isPlurConfigured check-after-home + expand CI matrix to Node 24 (#521) #523 — merged
2. 🔴 Merge PR fix(cli): per-session concurrency lock for hook-inject (#519) #520 (concurrency lock, closes P1 hook-inject cold-start too heavy (~20s BGE load, ~160MB RSS) — make per-invocation cheap #519) — CI green Node 20/22/24
3. 🔴 Merge PR feat(claw): migrate to before_prompt_build, sharpen ClawHub positioning #517 (ClawHub plugin) — CI green, ClawHub validation PASS
4. 🔴 Run publish:

   npx clawhub login
   npx clawhub package publish plur-ai/plur \
     --source-path packages/claw \
     --name openclaw-memory-plur \
     --owner plur-ai \
     --family code-plugin

5. 🔴 Post Show HN (weekday 8–10am ET) — draft in P0: Ship PLUR OpenClaw plugin to ClawHub — Supermemory beat us there #516 comments

🤖 heartbeat — cto.benchmark-clearance — 2026-07-11

[plur9]

Operator status — 2026-07-11

All pre-conditions are met. Blocked on human review gate.

What's ready

Item	Status
PR #520 — concurrency lock (P1 #519)	✅ CI green (Node 20/22/24), mergeable
PR #517 — ClawHub plugin positioning	✅ CI green, ClawHub validation PASS (0 findings)
Benchmark regression check	✅ Zero changes to minilm retrieval path since last run (see earlier comment)
awesome-openclaw PR #110	✅ Filed

Action needed from @-this-repo-owner

Branch protection requires a review before merge. Please:

1. Review + merge PR fix(cli): per-session concurrency lock for hook-inject (#519) #520 (2 min — adds a lock file guard, 2 test cases)
2. Review + merge PR feat(claw): migrate to before_prompt_build, sharpen ClawHub positioning #517 (2 min — hook rename + positioning text)
3. Run ClawHub publish:

   cd /path/to/plur
   npx clawhub login
   npx clawhub package publish plur-ai/plur \
     --source-path packages/claw \
     --name openclaw-memory-plur \
     --owner plur-ai \
     --family code-plugin

4. Post Show HN — draft is in earlier comment (weekday 8–10am ET)

Launch tweet thread is also ready in the earlier CMO comment.

🤖 heartbeat — ceo.operator — 2026-07-11

[plur9]

Ops merge-queue refresh — 2026-07-11

All 6 open PRs are now CI green (Node 20/22/24). PR #524 cleared CI overnight after #523 landed on main (the session-checkpoint flake that caused Node 20/22 failures is gone).

Complete merge sequence for the next human session

#	PR	Priority	Status
1	#520 — per-session concurrency lock	P1 (unblocks P0)	✅ CI green · LGTM
2	#517 — ClawHub plugin positioning	P0	✅ CI green · LGTM
3	Publish	P0	🔴 human gate (see publish command in earlier comment)
4	#524 — plur_status domain + created_after	P2	✅ CI green · 2× LGTM
5	#515 — benchmarks counter-positioning doc	P2	✅ CI green · LGTM
6	#513 — packs CLI subcommands	P2	✅ CI green · LGTM
7	#510 — onboarding recall transcript	P2	✅ CI green · LGTM

After step 3 (publish): post the Show HN (draft in CMO comment above) and the tweet thread.

🤖 heartbeat — ops.sprint-claim — 2026-07-11

[plur9]

Merge queue refresh — 2026-07-11 (post CTO tick)

PR #526 was filed since the last queue update. Adding it to the sequence.

Complete merge sequence

#	PR	Priority	Status
1	#520 — per-session concurrency lock (P1 #519)	P1	✅ CI green · LGTM
2	#517 — ClawHub plugin positioning (P0 #516)	P0	✅ CI green · ClawHub validate PASS · LGTM
3	Publish — npx clawhub package publish	P0	🔴 human gate
4	#524 — plur_status domain + created_after (#522)	P2	✅ CI green · LGTM
5	#515 — benchmarks counter-positioning doc	P2	✅ CI green · LGTM
6	#513 — packs CLI subcommands	P2	✅ CI green · LGTM
7	#510 — onboarding recall transcript (#508)	P2	✅ CI green · LGTM
8	#526 — RemoteStore AbortController covers body read (#525)	P3	✅ CI green · LGTM

All 8 PRs have clean CI (Node 20/22/24). No dependency ordering between 4–8; they can merge in any order after #517.

Publish command (step 3)

cd /path/to/plur
npx clawhub login
npx clawhub package publish plur-ai/plur \
  --source-path packages/claw \
  --name openclaw-memory-plur \
  --owner plur-ai \
  --family code-plugin

Post-publish

• Show HN — draft in earlier CMO comment
• Tweet thread — draft in earlier CMO comment
• mcp.directory — submission form ready in Distribution: submit PLUR to mcp.directory Claude Code memory roundup #518 (30 sec, human fills https://mcp.directory/submit)

🤖 heartbeat — ops.sprint-claim — 2026-07-11

[plur9]

Merge queue refresh — 2026-07-11 (evening)

PRs #528 and #529 opened since the 09:32 update. Adding to sequence.

Complete merge sequence

#	PR	Priority	Status
1	#520 — per-session concurrency lock (P1 #519)	P1	✅ CI green · LGTM
2	#517 — ClawHub plugin positioning (P0 #516)	P0	✅ CI green · ClawHub validate PASS · LGTM
3	Publish — npx clawhub package publish	P0	🔴 human gate
4	#524 — plur_status domain + created_after (#522)	P2	✅ CI green · LGTM
5	#515 — benchmarks counter-positioning doc	P2	✅ CI green · LGTM
6	#513 — packs CLI subcommands	P2	✅ CI green · LGTM
7	#510 — onboarding recall transcript (#508)	P2	✅ CI green · LGTM
8	#526 — RemoteStore AbortController covers body read (#525)	P3	✅ CI green · LGTM
9	#528 — Python SDK npx pin 0.10.1→0.12.0 + recall() CHANGELOG (#495)	P2	✅ CI green · LGTM
10	#529 — plur-langchain LangChain/LCEL adapter	P2	✅ CI green · LGTM

All 10 PRs are CI-green (Node 20/22/24). Merge #520 first — unblocks #517 and the ClawHub publish. 4–10 can merge in any order after step 3.

Publish command (step 3 — ClawHub)

npx clawhub login
npx clawhub package publish plur-ai/plur \
  --source-path packages/claw \
  --name openclaw-memory-plur \
  --owner plur-ai \
  --family code-plugin

After ClawHub publish

• Show HN — draft in earlier CMO comment on this issue
• Tweet thread — draft in earlier CMO comment on this issue
• mcp.directory — 30-sec form at https://mcp.directory/submit (details in Distribution: submit PLUR to mcp.directory Claude Code memory roundup #518)

Additional human-gated items (independent of ClawHub)

• PR fix(python): bump npx CLI pin 0.10.1→0.13.0, document recall() semantic change #528 + PyPI publish — cd packages/python && python3 -m build && twine upload dist/* (plur9 credentials)
• PR feat(langchain): add plur-langchain adapter for LangChain + LCEL #529 + PyPI Trusted Publishing setup — one-time: PyPI → plur-langchain → add plur-ai/plur/publish-langchain.yml as trusted publisher
• PR feat(mcp): packs install/list/uninstall CLI subcommands #513 — needs crtahlin or gregor to review (packs CLI, LGTM from CTO)

🤖 heartbeat — ops.merge-queue-refresh — 2026-07-11