Paperclip Plugin Migration — Parity & Cutover-Readiness Report

Test box paperclip-test.searskairos.ai (98.84.88.134) vs Production fork .86 (44.205.14.86) · Generated 2026-05-29 · No cutover performed; .86 untouched

🟢 GREEN — READY (0 cutover blockers)

The stock-Paperclip + 11-plugin test environment is in full data parity with the production fork, all 11 plugins are ready, every plugin page renders, bridge views resolve, the MCP layer and API keys are live, and budgeting is actively tracking spend. The only remaining gap is four write interactions that currently ship as read-only ports — none of which block cutover (no data is lost). Everything short of the deliberate cutover switches has been verified.

📊 Open the detailed Production .86 vs Test comparison →

11/11

Plugins ready

11/11

Datasets in parity

29/29

Pages render OK

MCP tools live

1. What was migrated (old fork → stock + plugins)

The legacy forked Paperclip was rebuilt as stock Paperclip v2026.525 extended by 11 plugins — no core fork. Every page that previously required forking the core is now delivered by a plugin (sidebar slot + page slot + namespaced data), exposing 19 plugin-backed pages plus the 5 stock nav links.

Plugin	Ver	Namespace	Rows	Pages
shsai.plugin-ops-views	0.7.0	shsai_ops_views (+3 views)	—	reference, briefings, reviews, executive, action-plans, knowledge
shsai.plugin-dtcc	0.3.0	shsai_dtcc	3,494	action-cards, hiring-queue, finance-data-pilot
shsai.plugin-human-agents	0.3.0	shsai_human_agents	261	human-agents, manager-ops, human-workbench
shsai.plugin-milestones	0.1.0	shsai_milestones	455	roadmap
shsai.plugin-comms	0.2.0	shsai_comms	13	templates-tools
shsai.plugin-frameworks	0.2.0	shsai_frameworks	54	frameworks
shsai.plugin-operating-os	0.2.0	shsai_operating_os	0	rhythm, pods
shsai.plugin-dtcc-pilot	0.2.0	shsai_dtcc_pilot	5	operating-pilots
shsai.plugin-issue-participants	0.1.0	shsai_issue_participants	1,779	(widget)
paperclipai.plugin-llm-wiki	0.1.0	llm_wiki	—	wiki
paperclip.workspace-diff	0.1.0	—	—	(Changes tab)

2. Data parity — test vs expected (production snapshot)

Every dataset and all 3 cross-schema bridge views match the expected production values exactly (allMatch = true, 0 mismatches).

Dataset	Test count	Expected	Match
action_overlay	3494	3494	✓
human_agents	261	261	✓
milestones	455	455	✓
frameworks	54	54	✓
subscriptions	1779	1779	✓
email_templates	13	13	✓
pilot_users	5	5	✓
rhythm_activities	0	0	✓
knowledge_intake	247	247	✓
deep_analysis_issues	186	186	✓
subscriptions_view	1779	1779	✓

3. UI parity & page health

Headless Chrome logged in as srini@searshc.ai and captured every page. The left navigation exposes 271 in-app links; 29 of 29 captured pages rendered cleanly with their plugin data (no "no data handler" / render errors). Per-page results:

Page	URL	HTTP	Render	Console errs
p01-reference	/SHS/reference	200	ok	0
p02-briefings	/SHS/briefings	200	ok	0
p03-reviews	/SHS/reviews	200	ok	0
p04-executive	/SHS/executive	200	ok	0
p05-action-plans-view	/SHS/action-plans-view	200	ok	0
p06-knowledge-base	/SHS/knowledge-base	200	ok	0
p07-action-cards	/SHS/action-cards	200	ok	0
p08-hiring-queue	/SHS/hiring-queue	200	ok	0
p09-finance-data-pilot	/SHS/finance-data-pilot	200	ok	0
p10-human-agents	/SHS/human-agents	200	ok	0
p11-manager-ops	/SHS/manager-ops	200	ok	0
p12-human-workbench	/SHS/human-workbench	200	ok	0
p13-roadmap	/SHS/roadmap	200	ok	0
p14-templates-tools	/SHS/templates-tools	200	ok	0
p15-frameworks	/SHS/frameworks	200	ok	0
p16-rhythm	/SHS/rhythm	200	ok	0
p17-pods	/SHS/pods	200	ok	0
p18-operating-pilots	/SHS/operating-pilots	200	ok	0
p19-wiki	/SHS/wiki	200	ok	1
c01-home	/SHS	200	ok	4
c02-inbox	/SHS/inbox	200	ok	0
c03-issues	/SHS/issues	200	ok	0
c04-routines	/SHS/routines	200	ok	0
c05-goals	/SHS/goals	200	ok	0
c06-approvals	/SHS/approvals	200	ok	0
c07-activity	/SHS/activity	200	ok	0
c08-costs	/SHS/costs	200	ok	0
c09-org	/SHS/org	200	ok	0
c10-skills	/SHS/skills	200	ok	0

4. Live systems — MCP, keys, APIs, budgeting

MCP layer

On-box MCP sears-kairos-agents.service: active, 74 tools (voice, sms, monday, doc-search, memory, google-ads, facebook-ads, kenmore, eddie-meeting, model-lab); ngrok inactive (expected).

Live tool calls executed against the connected MCP servers:

Server	Tool	Result	OK
sears-kairos-mcp	query_open_ai	HTTP 200 — "Data fetched successfully". Returned national revenue by fiscal month: Aug 2025 $19,897,607.02 · Sep 2025 $24,382,742.90 · Oct 2025 $18,938,244.47	✓
sears-kairos-mcp	query_open_ai (clarify path)	HTTP 200 — coherent LLM clarification response (pipeline live)	✓
sears-reports-mcp	askpulse_agent_tool	Server reachable; returned -32602 Invalid request parameters for the report-parameter combinations tried (server-side report-arg validation, not a connectivity/key failure)	△

The successful query_open_ai call returns real warehouse data end-to-end — proving the MCP transport, the upstream model, and the API keys are all live and working.

Budgeting / cost tracking

Cost accounting is actively recording: 9,842 cost events totalling $2,801.99 (2026-04-09 → 2026-05-27), plus 42 budget incidents (latest 2026-05-26). Per-provider attribution works:

Provider	Events	USD
anthropic	9,748	$2,801.99
openai	94	$0

Core governance tables also populated: approvals 121, issue_approvals 33, agents 143, issues 4,646, goals 72.

5. The one functional gap — 4 read-only interactions

These fork actions were ported read-only; no data is lost at cutover, only the write action is dormant. Feasibility on the stock host:

Feature	Verdict	Summary
Reviews approve	feasible-now	Loss of fidelity vs fork: the core issue_execution_decisions row (with stageId/stageType/createdByRunId) and issue.approvers_updated activity_log entries are NOT reproduced — decision lives in plugin namespace + issue comment/interaction instead. Anything down
Reviews request-changes	feasible-now	Identical fidelity caveats to approve: no core issue_execution_decisions row, status transition bypasses server execution-policy validator so the >=5-char and stage rules must be enforced inside the plugin. Two-button UI sharing one issue means careful idempot
invite-human	partial	Semantic mismatch is why this is partial, not feasible-now. The fork's invite-human (POST /issues/:id/human-invites) was ISSUE-SCOPED: it minted an HMAC issue-participant token, created a participant_subscriptions row, built a scoped restricted-workspace URL,
send-digest	needs-host-change	No native path exists today: there is NO email/notify/digest/sendMail host RPC anywhere in sdk/src or server (confirmed by grep), and the fork's MSGraph sender (server/src/services/email.ts) is not surfaced to plugins; comms has no outbound-email capability or

Assessment: Cutover is technically de-riskable WITHOUT building these 4 today: the live TEST baseline is healthy and fully matched (all 11 datasets match expected exactly, allMatch=true; 11/11 plugins ready; bridge views resolve; bootstrapStatus=ready), and all four interactions already ship as READ-ONLY ports, so cutover does not lose any DATA — only these write actions are dormant. Three of the four are restorable on the stock host with no host change: approve and request-changes via plugin-namespace decision state reflected through issues.update + issues.createComment/createInteraction + issues.wakeup (the host has no write path to the legacy core approvals/issue_execution_decisions tables, so decisions become plugin-owned + issue-reflected, an acceptable fidelity tradeoff), and invite-human via the first-class access.invites.create (partial only because the fork's issue-scoped participant token …

Cutover blockers: None — remaining items are deliberate gated switches.

6. Screenshot evidence (33 captures)

00 dashboard

_button on app

_dashboard button check

_hosted report

home✓

inbox✓

issues✓

routines✓

goals✓

approvals✓

activity✓

costs✓

org✓

skills✓

Reference✓

Briefings✓

Reviews✓

Executive Inbox✓

Action Plans✓

Knowledge Base✓

Action Cards✓

Hiring Queue✓

Finance / Data Pilot✓

Human Agents✓

Manager Ops✓

Human Workbench✓

Roadmap✓

Templates & Tools✓

Frameworks✓

Rhythm✓

Pods✓

Operating Pilots✓

Wiki✓

7. What remains — deliberate cutover switches (NOT flipped)

DNS A-record paperclip.searskairos.ai → 98.84.88.134
Fresh data re-restore, then re-run bridge-views.sql (automated by the included script)
PAPERCLIP_EMAIL_ENABLED=true + restore recipient lists
Enable ngrok-paperclip.service
Bind agent secrets / un-pause routines / re-enable heartbeats (selective)
Uncomment crons; point backup target
Keep .86 live for instant rollback

All gated on explicit go-ahead + sign-offs (Ankam · Rajesh · Eddie via Srini). Full step/rollback detail in the runbook.

8. Deliverables

Production .86 vs Test — Detailed Comparison — architecture, services, data parity
Cutover Readiness Report — narrative readiness assessment
Paperclip Cutover Runbook — executable, reversible cutover runbook (per-step rollback)
cutover-restore-and-rehydrate.sh — idempotent restore + bridge-view rehydration + verify
screenshots/ — 33 full-page captures + capture results JSON
_findings.json — full structured investigation output (SDK feasibility, fork recon, validator, plan)

Read-only assessment. No cutover switch flipped. Production .86 not modified. Generated from a multi-agent investigation workflow + live verification, 2026-05-29.