Skip to content

Amazon Agent Atlas

The Amazon knowledge your agent didn't have to earn.

Ungrounded, an agent guesses — it quotes last year's FBA fees, writes AMC SQL that won't compile, invents a dispute window. Atlas is the knowledge it checks first: 55,844 content indexes across Ads, Seller, and Vendor, with decision guidance applied during retrieval. Answers come back cited, in milliseconds.

corpus.index — 55,844 content indexes
amazon_ads
29,712
amazon_sellers
18,531
amazon_vendors
7,601
embed: BAAI/bge-small-en-v1.5
filter: domain · topic · doc_type
semantic retrieval · structured metadata · source-aware results

What Atlas knows

55,844 content indexes covering every Amazon system your team works with.

29,712
content indexes
amazon_ads

Amazon Ads & AMC

Sponsored Ads, DSP, AMC audiences, attribution, SQL templates, conversion paths, audience segmentation.

18,531
content indexes
amazon_sellers

Seller Central

FBA fulfillment, Brand Registry, listings, Buy Box, fees, inventory, shipping, returns, tax.

7,601
content indexes
amazon_vendors

Vendor Central

Compliance, chargebacks, ARA/ABA reports, A+ content, Net PPM, EDI/API integrations, catalog operations.

Why retrieval changes the agent

It looks up the answer before it answers.

?"lapsed customer re-engagement"
conversions_all_for_audiences.sql0.91
audience_iq_lapsed.md0.87
Semantic retrieval

Retrieves by meaning, not keyword

BAAI/bge-small-en-v1.5 embeddings find the right playbook even when your question doesn't match the docs verbatim.

domainamazon_ads
topicaudience_iq
doc_typesql_template
datasetconversions_for_audiences
Structured metadata

Filters before it answers

Every content index entry carries domain, topic, doc type, source path, dataset references, and code language tags.

"… DSP activation ~48h"
amc/audiences/activation.md#timing
"… min 500 users"
amc/audiences/limits.md#size
Source-aware results

Citations on every claim

Section paths and code language are preserved during indexing, so every recommendation links back to its source.

IF ORD_g ≥ 10 AND ACOS_g ≤ 0.9 × TGT
actionscale_winner
bid Δ+5–10% · clamp ±20%
reeval5d
Bounded actions

Rulebook, not improv

Scale ±5–10%, daily clamp ±20%, re-evaluate in 5 days. The agent consults the rulebook instead of guessing.

SELECT user_id
FROM conversions_for_audiences
WHERE event_subtype = 'shoppingCart'
compiled · 0 errors
AMC-fluent SQL

Compiles on first try

Knows conversions_for_audiences vs conversions_all, the right event_subtype codes, and the 14-day attribution wait.

LTSF endedJul 1 2024
dispute window30d
SnS in sandboxno
attribution wait14d
Operational guardrails

Knows what's current

Long-term storage fees ended Jul 2024. Chargeback dispute window is 30 days. Sandbox doesn't populate SnS.

AMC SQL playbooks

Retrieved AMC SQL. Compiles on the first run.

AMC gives you cross-channel attribution, audience overlap, and customer journey analysis. The SQL is the wall most teams stall against. Atlas indexes the playbooks Amazon publishes — and the operational details Amazon doesn't.

AMC · Audience IQ

"Build me a DSP audience of people who added one of our hero ASINs to cart but didn't buy in the last 30 days."

Retrieves the cart-abandonment Audience IQ, knows it must use conversions_for_audiences (the only variant that allows SELECT user_id), and produces a query that compiles on the first try.

Also surfaces: AMC audiences require min 500 / max 500k users, DSP activation takes ~48h, and shoppingCart can be swapped for detailPageView to reach further upstream.
amc.sql
WITH asins (asin) AS (
  VALUES ('B0XXXXXXX1'), ('B0XXXXXXX2')
),
purchase AS (
  SELECT user_id, MAX(event_dt_utc) AS purchase_dt_max
  FROM conversions_for_audiences
  WHERE event_subtype = 'order'
    AND tracked_item IN (SELECT asin FROM asins)
  GROUP BY 1
),
atc AS (
  SELECT user_id, MAX(event_dt_utc) AS atc_dt_max
  FROM conversions_for_audiences
  WHERE event_subtype = 'shoppingCart'
    AND tracked_item IN (SELECT asin FROM asins)
  GROUP BY 1
)
SELECT atc.user_id
FROM atc
LEFT JOIN purchase ON atc.user_id = purchase.user_id
WHERE atc_dt_max > purchase_dt_max
   OR purchase_dt_max IS NULL;
AMC · Subscribe & Save

"How much more do my Subscribe & Save customers spend compared to one-off buyers?"

Knows that Feb 2024 extended FSI to include repeatSnSOrder alongside snsSubscription and firstSnSOrder, and returns a working comparison query.

Flags the constraints that aren't in the SQL: query ≥3 months for SnS cadence, narrow with an ASIN filter for performance, and skip Sandbox — sns_subscription_id isn't populated there.
amc.sql
SELECT
  CASE
    WHEN event_subtype IN (
      'snsSubscription','firstSnSOrder','repeatSnSOrder'
    ) THEN 'subscriber'
    ELSE 'non-subscriber'
  END AS user_type,
  COUNT(DISTINCT user_id) AS users_that_purchased,
  SUM(total_product_sales) AS total_sales,
  SUM(total_product_sales) /
    COUNT(DISTINCT user_id) AS average_spend
FROM conversions_all
GROUP BY 1;
AMC · Path-to-Conversion

"What's the actual journey our customers take across DSP, SP, SB, and SD before they convert?"

Pulls Customer Journey Analytics, identifies the three required tables (dsp_impressions, sponsored_ads_traffic, amazon_attributed_events_by_traffic_time), and outputs a result shaped for a Sankey diagram.

Knows to wait 14 days past the query end date for attribution to close, and points you at the newer Sponsored Ads & DSP Overlap IQ that handles 2/3/4-way overlap instead of the deprecated version.
amc.sql
-- Path-to-Conversion grouped by campaign category
-- Joins DSP impressions, sponsored ads traffic,
-- and attributed events on user_id within window.
SELECT
  path_signature,
  campaign_category_path,
  COUNT(DISTINCT user_id) AS converters,
  SUM(total_product_sales) AS sales
FROM customer_journey_paths
GROUP BY 1, 2
ORDER BY converters DESC;

Beyond AMC

FBA fees. Chargebacks. Bid math. The unglamorous Amazon work, handled.

Vendor chargeback disputes

Pulls the exact 30-day dispute window, the second-dispute path, and the Smart Aggregation Processing 7-day preview cycle. Drafts the dispute in the evidence pattern Amazon's reviewers actually look for.

FBA inventory & storage

Knows long-term storage fees ended July 1, 2024. Quotes the active fees — monthly storage, aged-inventory surcharge, capacity limits in cubic feet — instead of the 2022 schedule.

Brand Registry conflicts

"Trademark already enrolled" → retrieves the specific resolution path: contact the existing Administrator via the application flow, or file via the FAQ's "use this form" mechanism if inactive.

Bid decisions, not bid vibes

Scale-winner fires only when ORD_g ≥ 10 AND ACOS_g ≤ 0.9 × TARGET_ACOS. Increase 5–10%, daily clamp ±20%, re-evaluate in 5 days. Conflict order: invalid → thin-data → safety → pull-back → scale.

Wired into a workflow

Retrieved and adapted, not generated.

Atlas is the retrieval layer beneath agents that already do work — pulling SP-API reports, querying AMC, drafting support replies, auditing campaigns, building audiences, planning inventory.

01

Your team asks in plain language

"Build me a high-value lookalike for Prime Day."

02

Agent searches the corpus semantically

Retrieves the AMC Lookalike Audiences for Promotional Events playbook, the high-value seed query template, and the 500–500k size constraints.

03

Agent grounds the response

Compilable seed query, named thresholds, the testing approach the playbook actually recommends, and the 48-hour DSP activation lag.

04

Your team runs it

Works the first time — because the answer wasn't generated, it was retrieved and adapted.

Frequently asked questions

Common questions.

What is Amazon Agent Atlas?
A retrieval layer your agent checks before it answers. It indexes Amazon operating knowledge across Ads, Seller Central, and Vendor Central, applies decision guidance during retrieval, and returns cited answers in milliseconds.
Why does an agent need this if the model already knows Amazon?
Because an ungrounded model guesses. It quotes last year's FBA fees, writes AMC SQL that will not compile, and invents dispute windows. Atlas makes the agent look the answer up instead of recalling it.
What does Atlas actually cover?
Amazon Ads and AMC, Seller Central, and Vendor Central. Decision rules, sufficiency gates, conflict resolution, action enums, and category style guidance are applied within that knowledge rather than presented as a separate catalog.
Does Atlas hold my account data?
No. Atlas is operating knowledge about how Amazon works, not a copy of your business. Live figures from your own accounts come through the MCP servers and Amazon Agent Flow.
Can I see where an answer came from?
Yes. Retrieval is source-aware and answers come back cited, so you can trace a claim to the document it was drawn from rather than trusting the model's memory.

The bottom line

Your hard-won Amazon knowledge, finally something AI can use.

Atlas grounds every answer in your data and your playbook — every indexed result traceable to its source, every recommendation cited. The edge you built stays yours, not the model's.