Amazon Seller MCP Server Docs & AI Setup

Connect Claude - two steps, no keys

Add the connector - open the AMZ Vault listing in Claude's connector directory and click Connect (also findable in Claude under Settings → Connectors). Prefer manual setup? Add custom connector with our endpoint: https://www.amz-vault.com/mcp - same browser sign-in.
Sign in when prompted - Claude opens the AMZ Vault consent screen; approve, and the tools appear in every conversation.

AMZ Vault is published on the official MCP Registry as io.github.TheBurdz/amz-vault - the canonical, protocol-maintained record of the server - and listed in Claude's connector directory, Glama, AllMCPs, and other MCP directories.

Claude Code & other MCP clients

Clients that authenticate with a header instead of a browser sign-in (Claude Code in the terminal, custom MCP integrations) use a personal access token: in AMZ Vault, Settings → AI Connector → Generate token - scoped to your account, revocable any time, shown with the exact connect command. Keep the two paths separate: connecting through the Claude directory never needs a token.

Trouble connecting? support@amz-vault.com answers fast.

How writes work

stage_* tools save a proposal - a diff of exactly what would change - and nothing more. The change reaches Amazon only when confirm_staged_changes is called, and your AI client shows you the full change set before that happens. Read-only team roles can't stage or confirm at all: read-only means read-only, everywhere, including the AI connection.

Your desk (Today, Decisions, departments) (3)

department_report read One department page as data: the signal ledger, pulse bars, open decisions (with item ids for desk_decisions), what moved in the last 14 days versus the 14 before, the AI team's coverage, how to run it yourself, and what

desk_decisions read The Decisions queue exactly as the owner's desk ranks it: staged proposals (from seats or chat), optimizer batches and platform signals (budget caps, listing issues, reorders, COGS holes, A+ gaps, account health, refund

desk_signal_action write Snooze a signal tile for N days, dismiss it until its count grows, or clear every snooze - the same personal view state the desk's Snooze and Dismiss buttons write

Account & discovery (12)

account_feedback write Feedback & support channel to the AMZ Vault team (NOT an Amazon write): action=submit opens a thread (subject, body, category bug|idea|question|other), action=list shows this account's threads WITH team replies, action=r

account_fg_clicks read Field-guide daily-series click stats (ADMIN-ONLY): first-party clicks on the /fg social short links, crawler-filtered, split by post (series day + chapter), source (linkedin/facebook/social), guide chapter, and calendar

account_health_check read Per-connection onboarding + sync health across ALL tenants (SUPER-ADMIN ONLY)

account_help read Search and read AMZ Vault's Help Center - the same articles users see at /help: setup, connections and data semantics, per-page platform guides, AI/MCP setup, workforce, business playbooks, and troubleshooting

account_management_guide read The operator's brand & PPC decision framework: margin spine (PPD + margin bands), product postures, campaign structure/bidding, harvesting/negation logic, guardrails and decision triggers

account_notice_draft write Queue a one-off notice email to a user for admin review (ADMIN-ONLY)

account_ops_events read Operational health feed (ADMIN-ONLY): deduped errors / warnings / API-throttle events from the worker and web services with occurrence counts, plus failed SP-API sync runs and failed scheduled tasks

account_profiles read List the advertising profiles (PPC accounts) bridged to this token's seller connections.

account_report read THE canonical account overview — ONE call: period cards + settlement waterfall + freshness caveats for the money datasets, with quote_rules

account_sellers read List the Amazon seller connections this token can access, with their marketplaces and currencies.

account_storage read Database size by table (ADMIN-ONLY): total database size, the biggest tables with heap / index / TOAST split, live and dead row estimates, and whether any retention prunes each table

account_support_access write Grant or revoke the AMZ Vault support team's READ-ONLY view of this account: action=grant lets support see what the user sees for 7 days (auto-expires, they can change nothing, and every view notifies the user); action=r

Data warehouse (3)

data_describe read Without arguments: list every warehouse dataset available to data_query

data_freshness read Per-dataset sync freshness, row counts, date coverage and backfill progress — use to caveat answers correctly

data_query read Generic query over any warehouse dataset (see data_describe): column projection, filters (eq/neq/gt/gte/lt/lte/in/contains), date bounds, ordering, limit ≤1000 + offset — page with offset until a short page comes back to

P&L / profitability (9)

pnl_breakdown read Settlement-basis account waterfall: revenue → refunds → COGS → gross profit → Amazon fees → contribution margin → advertising → net profit → expenses → net profit.

pnl_by_asin read Per-ASIN P&L rows for a date window (optionally one ASIN): sales, units, fees, COGS, margins, TACoS, stock.

pnl_expenses read Custom operating expenses (subscriptions, VA, freight, 3PL…): scope, recurrence, amounts.

pnl_mcf_summary read Multi-Channel Fulfillment (off-Amazon) totals: units, orders, fulfillment fees and COGS for a window.

pnl_period_cards read The dashboard mini-P&L cards (Today / Yesterday / MTD / This-month forecast / Last month).

pnl_refunds read Refund components for a window: net refund amount, count, and itemized lines (reimbursements, restocking).

pnl_sb_attribution read Per-ASIN Sponsored Brands cost attribution (the sales-weighted split of campaign-level SB spend).

pnl_settlement_audit read Settlement fee audit: every distinct fee combination Amazon charged, classified into P&L buckets — verify nothing is misclassified or missed.

pnl_summary read Profit & Loss summary for a date window: account tiles (sales, fees, COGS, ad spend, gross profit, contribution margin, net profit, TACoS) plus the top product rows

Products & inventory (5)

inventory_health read FBA fulfillable/inbound/reserved units, aged inventory, restock recommendations and AWD pool per SKU

inventory_shipment_tracking read Live supply-chain tracking (SP-API Tracking, US store only): milestones and estimates for a shipment moving through Amazon - pass ONE identifier (aftn, acsin, container_number, house_bill_of_lading_number, id, or carrier

product_cogs read Current per-SKU landed costs (COGS) and coverage summary.

product_fee_preview read Amazon's own estimated fees per SKU for the coming weeks (referral, fulfillment, storage).

product_library read The canonical product catalog with breakeven ACoS and effective BE margin targets per product family.

Brand analytics (6)

brand_market_basket read What customers buy together with each ASIN (bought-with pairs and combination shares).

brand_repeat_purchase read Customer loyalty per ASIN: repeat buyer share and repeat revenue by week.

brand_search_catalog read Per-ASIN search funnel totals across ALL queries (query-less SQP): impressions, clicks, cart adds, purchases, median prices.

brand_sqp read Search Query Performance: organic search funnel per (ASIN, query) — impressions/clicks/cart-adds/purchases with shares

brand_store_analytics read Brand Store performance (Stores Analytics API): view='insights' = store traffic/sales by DATE / PAGE / SOURCE / STORE / TAG (visits, visitors, views, sales, units, orders; NEW_TO_STORE/DWELL_TIME/BOUNCE_RATE need dimensi

brand_top_search_terms read Marketplace-wide top search terms with the top-3 clicked ASINs and their click/conversion shares

Keyword intelligence (3)

keyword_finder read Keyword Finder (reverse-ASIN / multi-input lookup): every keyword phrase the input ASINs are known for, merged from first-party data — SQP (own ASINs, TRUE Amazon search volume), Top Search Terms (ANY marketplace ASIN th

keyword_opportunities read KW research FRONT DOOR - start here for any keyword question about a product (new or established)

keyword_research read Keyword Research (marketplace term universe): search the FULL Top Search Terms store by phrase - not anchored to any ASIN

PPC — read (35)

ppc_account_summary read Get account-level aggregate metrics for a date range: total spend, sales, ACoS, ROAS, clicks, impressions, purchases.

ppc_ad_groups read Query ad group performance data within campaigns

ppc_ad_type_breakdown read Get performance comparison across ad types (SP vs SB vs SD)

ppc_advertised_products read Query product/ASIN-level performance (SP+SD ad grain, merged per ASIN with a per-program split; SB per-ASIN spend exists only in the blended P&L)

ppc_asset_upload_link read Mint a 1-hour browser-upload link so the USER can hand over a local file (chat attachments cannot reach the server)

ppc_bid_recommendations read Amazon's suggested bid + range per keyword for a SPECIFIC campaign (routed by campaign type)

ppc_budget_alerts read Find campaigns approaching or exceeding their daily budget - SAME-DAY where Marketing Stream data flows (signal_source=stream rows are minutes-fresh pushes; capped_since_utc shows when a campaign hit 100% today, separati

ppc_campaign_create_guide read The verified contract for creating campaigns via stage_campaign_create: per-format required fields, goals, landing-page rules, asset sourcing, safety model, and post-create lifecycle

ppc_campaign_details read Get detailed data for specific SP campaigns including placements, advertised ASINs, placement modifiers, and top keywords/targets

ppc_campaign_metadata read Campaign configuration and user metadata: bidding strategy, placement settings, goal/KPI, tags, notes.

ppc_campaign_search read Lightweight campaign name lookup

ppc_campaigns read Query campaign performance data with optional filters

ppc_change_history read Get recent campaign change history showing what changed, when, old values, and new values

ppc_conversion_funnel read Get conversion funnel data: Impressions → Clicks → Orders with rates between each stage

ppc_creatives read SB ad creatives at ad grain (headline, brand logo, custom images, video asset ids, ASINs, landing page, creative status) plus SD creatives; include_assets=true adds the asset-library inventory.

ppc_daily_trend read Get day-by-day metrics for trend analysis

ppc_dayparting_analytics read Dayparting schedule AND real hour-of-day performance

ppc_describe read Without arguments: catalog of the raw PPC datasets behind ppc_query (daily report tables at campaign/ad-group/target/search-term/placement/ASIN grain for SP+SB+SD, plus current-state snapshots: campaign_meta, portfolios,

ppc_harvesting_candidates read Find converting search terms eligible for promotion to manual campaigns

ppc_keywords read Query keyword/target bid performance across Sponsored Products AND Sponsored Brands (rows carry ad_type SP/SB; SD targeting expressions live on ppc_query dataset sd_targets)

ppc_match_type_breakdown read Get performance breakdown by keyword match type (Broad, Phrase, Exact, etc.)

ppc_negation_candidates read Find non-converting search terms with significant spend — candidates for negative keyword addition

ppc_negatives read Current negative keywords and negative product targets (campaign- and ad-group-level), with state and source

ppc_new_to_brand read New-to-brand purchases/sales and NTB% per campaign (SB + SD, from standard report columns — no AMC needed)

ppc_optimization_suggestions read Analyze campaigns and categorize into top performers (at or below target ACoS), running hot (significantly over target), and wasted spend (spending with no sales)

ppc_period_comparison read Compare current period metrics to the previous period of equal length

ppc_placement_modifiers read Current placement bid modifiers per campaign (top of search, product pages, rest of search).

ppc_placements read Get performance breakdown by placement type (Top of Search, Product Pages, Rest of Search) from SP placement data

ppc_portfolios read Get portfolio-level performance rollup

ppc_product_family read Look up parent/child product hierarchy from the product library (Category Listing report)

ppc_product_report read THE canonical product/family report — ONE call returning every view labeled: family resolution, asin_accurate ad totals (SP+SD ad grain — the numbers to quote for ad spend/ACoS; SB per-ASIN spend is in the blended pnl on

ppc_prompts read Sponsored Prompts: the conversational ads Amazon shows inside Alexa for Shopping (formerly Rufus) and on search and detail pages, with the prompt text Amazon wrote, per prompt x ad x ad group over the window (SP and SB)

ppc_query read FULL-fidelity row-level query over any raw PPC dataset (see ppc_describe): column projection, filters (eq/neq/gt/gte/lt/lte/in/contains), date bounds, ordering, limit ≤1000 + offset — page with offset until a short page

ppc_recommendations read Amazon's own consolidated recommendations (unified Recommendations API): bid/budget/bidding-strategy changes, new keywords, negatives, targets, placement and state suggestions across SP/SB/SD

ppc_search_terms read Query search term performance data

Automation (6)

automation_activity read THE approval inbox for V2 rule automation — the ONLY surface for pending automation approvals (MinMax/AOE/scheduled-task tools are separate systems, not this inbox)

automation_dayparting read Dayparting, READ-ONLY: action=list shows schedules + assignments (mutations live in stage_dayparting_schedule).

automation_minmax_results read Recent MinMax budget-analysis runs: scope, top performers, campaigns running hot

automation_recent_tasks read Get the list of scheduled automation tasks with their status, cadence, and last execution results.

automation_rules read V2 rule-based automation, READ-ONLY: action=list|get|dry_run (mutations live in stage_automation_rule)

automation_scheduled_tasks read Scheduled tasks: type, cadence, enabled state, last and next run times

A+ content (1)

aplus_document read Fetch ONE A+ content document LIVE, including its full contentModuleList — the copy source for edits (the aplus_content dataset holds metadata only)

Staged writes (nothing reaches Amazon until you confirm) (31)

cancel_staged_changes staged write Cancel a pending proposal without executing it.

confirm_staged_changes write EXECUTE a pending proposal — call ONLY after you have rendered the full diff to the user as text (entities by name, current → proposed, reason) and they agreed

stage_ad_group_create staged write Create a new SP ad group in an existing campaign (name + default bid; created ENABLED)

stage_ad_group_update staged write Rename an ad group (SP/SB) or change its default bid (SP only — SB ad groups have no default bid)

stage_aplus_update staged write Create or update an A+ (EBC/Premium) content document, optionally attach ASINs and submit for Amazon approval

stage_asset_upload staged write Upload an IMAGE or VIDEO into the Creative Asset Library — from a public source_url, or from a staged_file_id (browser upload via ppc_asset_upload_link; images only)

stage_automation_review staged write Stage a decision on automation approval-inbox items (see automation_activity for the inbox itself): action=approve|reject|undo, each takes log_ids (max 100) + optional reason, and stages a proposal — NOTHING happens unti

stage_bid_changes staged write Propose bid changes on SP + SB + SD keywords/targets matching filters (ad_type narrows to one program)

stage_budget_changes staged write Propose SP campaign daily-budget changes

stage_campaign_create staged write Create Sponsored Products, Sponsored Brands or Sponsored Display campaigns end-to-end

stage_campaign_tags staged write Tag/annotate a campaign in AMZ Vault (add_tags / remove_tags / note) — pure metadata, nothing is sent to Amazon

stage_campaign_update staged write Update campaigns: name, budget_amount, budget_type (LIFETIME is SB-only), bidding_strategy, start_date / end_date (YYYY-MM-DD; null end_date clears it), portfolio_id (null unassigns), bid_optimization (SB-only bool), and

stage_cogs_update staged write Propose setting per-SKU landed costs (COGS), time-versioned

stage_competitor_watchlist staged write Propose adding/removing competitor ASINs on the review-insights watchlist

stage_entity_status staged write Pause/enable/archive specific entities BY ID: ad groups, product ads, keywords, targets, or campaigns (complements the filter-based stage_status_change)

stage_listing_update staged write Update LIVE Amazon listing content: title, item_highlights (the 125-char searchable companion field - over-cap is refused at staging; Amazon rejects it with issue 90225), bullet_points (full replacement list), descriptio

stage_margin_target staged write Propose a manual BE (breakeven) margin target for a product family (null be_margin_pct clears it so the computed breakeven applies)

stage_negatives staged write Propose adding negative keywords / negative ASIN targets

stage_negatives_remove staged write REMOVE (archive) negative keywords / negative product targets by their local ids from ppc_negatives — the other half of negation management

stage_po_upsert staged write Propose creating or updating purchase orders (matched by po_number per connection): supplier_name (created if new), status/milestone dates, ship mode, freight and duty, payments ([{label, amount, due_date, paid}]), notes

stage_portfolio_create staged write Create a new portfolio (no budget cap; set budgets after via stage_portfolio_update, incl

stage_portfolio_update staged write Update a portfolio: name, budget_amount, budget_policy (DATE_RANGE/MONTHLY_RECURRING/"remove"), budget_start_date, budget_end_date

stage_product_ad_add staged write Add advertised products to an existing SP or SD ad group (campaign_type, default SP) — created ENABLED

stage_product_edit staged write Propose editing one Product Library field (locks it against future syncs)

stage_status_change staged write Propose enabling/pausing campaigns matching filters

stage_supplier_upsert staged write Propose creating or updating suppliers (matched by name per connection): contacts, payment terms, and the production/transit lead times that power the reorder advisor

stage_targeting_add staged write Add keywords, ASIN product targets, and/or category targets to an EXISTING campaign + ad group (SP or SB) — e.g

staged_changes_history read Approved decisions and what they did - the verification loop

staged_changes_pending staged write List this account's pending (unconfirmed) proposals

Other (10)

app_today read Use this when the user asks for their Today desk, their dashboard, "what needs me today", what to approve, or an overview of the brand right now

forecast_reorder_report read Reorder advisor for one seller connection: per-ASIN runout dates from the latest forecast (counting FBA + AWD + inbound + open purchase orders), lead-time-aware reorder-by dates, suggested units, and an ADVISORY affordab

forecast_report read THE 12-month per-ASIN unit forecast (nightly engine): brand totals vs last year, seasonality index, diagnostics incl

growth_fact_base read The brand's quarter fact base - the deterministic facts a quarterly growth plan is written from, computed by AMZ Vault from the warehouse (no report pulls needed): account P&L for the trailing 90 days vs prior quarter, l

growth_plan read The brand's quarterly growth plan as filed in AMZ Vault: window, north star, scorecard (baseline -> week 4/8/13 targets), tasks (P0-P3, lane, week, end goal, checkpoints, status), risks, progress (plan week, phase, done

growth_plan_upsert write File or update the quarterly growth plan for one brand (AMZ Vault tables only - nothing reaches Amazon)

lab_term_gate read Intelligence Lab (SUPER-ADMIN ONLY): the confidence-gated negation/harvest engine run in SHADOW on one advertising profile

workforce_check_in write Check in with the AMZ Vault AI Workforce and receive the next worker packet

workforce_report write Submit the results of a workforce packet

workforce_status read Read-only view of the AI Workforce: hired roles, queue depth, staged changes awaiting dashboard approval, usage advisor, and recent runs

Docs - the AI connection