Five tools, five answers
The directory says a laptop was seen yesterday, the CMDB says it was retired, the endpoint tool has never heard of it. Which one do you report?
Fleet asset, audit & reconciliation
Turn scattered IT data into one record you can trust.
FAAR merges exports from your directory, CMDB, endpoint and service-desk tools into a single record per device, scores every device against your standards, and tests 61 possible root causes against the evidence.
- 420 devices reconciled in the sample workspace
- 88 fields tracked per device
- 24 of 61 root causes validated on the sample
- 100+ indicators to track
Bring data in fromCSV exportsREST APIs (JSON)Files and object storageDirectory, CMDB, endpoint and service-desk exports
Platform
One governed record for every device
FAAR brings your directory, CMDB, endpoint and service-desk data together, scores each device against your standards and shows what to fix first.
Why fleet data disagrees with itself
Every tool sees part of the picture. Put them side by side and they contradict each other - and nobody can say which one is right.
Errors that hide in plain sight
On the bundled sample alone FAAR catches 24 values recorded in the wrong unit, impossible dates, and 8 duplicate service-desk tickets - the kind of thing a merged spreadsheet quietly keeps.
Reports nobody trusts
Numbers merged by hand go stale the day they are sent, and no one can show where a value came from. FAAR keeps the source of every value.
Everything between a raw export and a decision
One workspace to connect data, reconcile it, understand it and act on it.
Connect
Bring data in without a project
Upload a CSV, call a REST API or read files from a folder or object storage. Credentials are held as references and never stored with the connection. Every sync records its health and any change in the file's columns, so a silent format change doesn't silently break your numbers.
- CSV upload, REST API (with paging and auth) and object storage or files by URL
- Health and schema-change tracking for every source
- A searchable inventory of sources, datasets and dashboards with owners and tags
Reconcile
One golden record per device, with receipts
Duplicates and conflicts are resolved field by field using a trust order you control, and every value in the record keeps the name of the source it came from. You never have to wonder where a number came from.
- A trust order per field: which source wins, and in what order
- Units and dates normalised; impossible values flagged
- Full traceability from every value back to its source
Design
Design the pipeline on a canvas. Read it as Python.
Drag sources, cleaning steps, quality rules and joins onto a canvas that ends in the golden record. The same pipeline is a readable Python script you can edit, review and keep in version control - no SQL anywhere.
- Quality rules that warn, drop, quarantine or stop a run
- A run history with per-step row counts and an event log
- Python and canvas stay in step: edit either one
Understand
Test 61 possible root causes against your data
Instead of guessing why boot times are slow or a site is struggling, FAAR tests a library of 61 hypotheses against your devices and ranks each as validated, invalidated or inconclusive - with the evidence and the sample size behind it.
- Every device scored pass or fail against your hardware and performance standards
- Cohorts by site, persona and age to see where a problem concentrates
- Add or change a standard, check or hypothesis without writing code
Act
Ask, track goals, and hand over the audit
Ask questions in plain language, track KPIs and goals on dashboards, and export the audit report as a Word or PowerPoint document for the people who decide.
- An assistant that can see the workspace - with your own AI key, or built-in rules without one
- Charts, goals and dashboards you build from the same data
- Audit report export to Word and PowerPoint
How it works
Four steps from scattered exports to answers you can act on.
- Your toolsDirectoryCMDBEndpointService deskFiles and APIs
- HYHyperconnect
Every source, one intake
- RReconcile
One record per device
- OOptimize
Scored and ranked
- XXecuteAssistantGoalsAudit report
Hyperconnect
Pull in data from every tool you already use - CSV files, REST APIs, files in storage - with no manual merging.
Reconcile
Catch duplicates, conflicts, wrong units and impossible dates and merge them into one record per device. You decide which tool wins for each detail.
Optimize
Score every device against your own standards and rank the likely root causes, so you know what to fix first.
Xecute
Ask the assistant, track goals and hand over the audit - so an answer turns into an action.
Built to explain, not just to count
Independent of your tools
Any CSV, API or file. You name the sources and map their columns. Nothing is tied to one vendor.
It explains
Root causes are tested against your data, with the evidence and the sample size.
Every value has a source
One record per device that never forgets where each value came from.
Open, and yours to run
Readable Python, a REST API and your own data model, on a machine you host.
Integrations
Generated from the connector list in the product, so what appears here is what exists. Systems that are planned are shown separately and are not available.
8 shown
CSV upload
SupportedUpload a CSV export from any system, then map its columns to canonical fields.
Limits
- A manual upload - there is no schedule or refresh; upload again to update.
- CSV only (UTF-8).
REST API (JSON)
SupportedRead records from an HTTP endpoint that returns JSON or JSON Lines.
Limits
- Read-only: it never writes to the source.
- Pagination styles: page number, offset, cursor and Link-header only (@odata.nextLink is not followed).
- No OAuth flows - use a token or key held in a secret.
- Server-side calls are guarded: link-local / cloud-metadata addresses are always refused and FAAR_CONNECTOR_BLOCK_PRIVATE=1 also refuses private ranges; redirects are re-checked (max 3).
- Each sync re-reads everything; there is no incremental cursor yet.
Object storage / files by URL
LimitedRead CSV, JSON, JSON Lines, Parquet or Excel files from a folder, a URL, or cloud object storage.
Limits
- file:// folders and http(s) URLs are verified.
- s3://, gs:// and abfs:// need the optional fsspec drivers (fsspec plus s3fs / gcsfs / adlfs) and have NOT been verified against live storage in this release.
- Excel (xlsx) needs the optional openpyxl package; only the first sheet is read.
- Read-only; no change tracking - every sync reads the matching files again.
- Several files are concatenated; all must share a schema (differences are reported).
- An administrator can confine file:// reads to chosen folders with FAAR_FILE_ROOTS; http(s) URLs go through the SSRF guard and are capped at 200 MB.
SQL database (read-only)
LimitedRead one table, or the result of one SELECT query, from SQLite, PostgreSQL, MySQL, SQL Server, Snowflake, BigQuery or Oracle. Read-only and row-limited.
Limits
- Read-only: one SELECT (or WITH ... SELECT) per source; writes, DDL, procedures and multiple statements are refused, the session is opened read-only where the database supports it, and the result is row-limited.
- Only SQLite is verified in this release. PostgreSQL, MySQL, SQL Server, Snowflake, BigQuery and Oracle need their optional drivers and have NOT been run against a live server here.
- SQL Server, Snowflake and BigQuery have no read-only session mode - give FAAR an account that can only read.
- SQLite files must be inside the folder the administrator names in FAAR_SQLITE_DIR.
- BigQuery uses the server's own Google credentials (no password field is used).
- Every value is read as text; the host is checked by the network guard but the driver connects by name.
- Each sync re-reads the table or query; there is no incremental cursor yet.
Kafka topic (batch)
LimitedRead a bounded batch of JSON messages from a Kafka topic - the last N, from the beginning, or from a time - without committing offsets.
Limits
- A bounded batch, not a continuous feed: each sync reads at most max_messages (hard cap 100,000) up to the topic's end at the moment the read starts.
- Read-only: offsets are never committed, no consumer group is joined, nothing is produced.
- JSON messages only; messages that aren't JSON are counted as unreadable. No Avro / Protobuf / schema registry.
- Needs the optional kafka-python package. Not verified against a live broker in this release (tested with a fake consumer).
- The broker addresses are checked by the network guard, but the driver connects by name.
Google Sheets (shared by link)
LimitedRead one tab of a Google Sheet that is shared as 'Anyone with the link can view', through its CSV export.
Limits
- Public-by-link sheets only: there is no OAuth and no Google sign-in. A private sheet is reported, not read.
- One tab per connection; values are read as text exactly as the CSV export shows them.
- The request goes through the SSRF-guarded HTTP path (redirects re-checked, 200 MB cap).
- Not verified against live Google in this release (tested against a local server shaped like the export).
Databricks tables
PlannedSalesforce
PlannedNothing matches. Clear the search or the filters.
Made for the people who own the fleet
The same reconciled data, useful from four seats.
IT asset managers
- See every device once, with the source of each detail
- Spot warranty, refresh-cycle and firmware gaps against your own standards
- Know which records disagree before an audit does
Security and compliance
- Score every device pass or fail against the standards you set
- See outdated firmware and unmanaged devices in one list
- Export the audit report as a Word or PowerPoint document
Service desk and support
- See where incidents cluster by site, model or firmware
- Test whether a hardware model or site really is the cause
- Deduplicate tickets and match them to the right device
IT leadership
- Track goals and KPIs on dashboards built from trusted data
- See which sites and cohorts lag, and by how much
- Ask questions in plain language
Fleet data, explained
Short answers to the questions people search for before they buy anything.
What is a golden record?
A golden record is the single, best version of what you know about a device (or any entity), assembled from several systems that each hold part of it. Instead of five conflicting rows you keep one, and for every value you can say which system it came from.
What is data reconciliation?
Reconciliation is matching records that describe the same thing across systems and resolving the differences. It needs an identity rule (what makes two rows the same device) and a trust order (which system wins for which field). FAAR lets you set both without code.
What is schema drift?
Schema drift is when a source quietly changes shape - a column is renamed, added, removed or changes type. Reports built on it keep running but start to mislead. FAAR compares each sync with the previous one and shows what changed and what depends on it.
How does root-cause testing work?
Each hypothesis is a testable statement, such as "slow boot clusters on one hardware model". FAAR runs the statistical test against your devices and marks it validated, invalidated or inconclusive, with the sample size and the evidence, so you can challenge the result rather than take it on trust.
What is a conformity score?
A conformity score is the share of devices that meet a standard you define - enough RAM, a current BIOS, within warranty, boot time under a limit. Each device passes or fails each standard, and scores roll up by site, cohort or fleet.
Trust and control
Built so you can check its work.
Every value has a source
Each field in a golden record keeps the name of the system it came from, and the rules that decided it are yours to edit.
Your organization, separated
Each organization has its own catalog and settings. Roles - viewer, editor, admin, owner - control who can change what, and API tokens are scoped to a person.
Credentials stay out of the data
Connections hold references to secrets, not the secrets. Values are read from the server's environment when a connector runs.
AI is optional
The assistant works with built-in rules out of the box. Add your own AI key for free-form questions; nothing is sent to an AI provider without one.
English and French, light and dark
The whole interface is available in both languages and both themes, with text contrast checked against WCAG AA.
Frequently asked questions
What is FAAR?
FAAR (Fleet Asset, Audit & Reconciliation) merges data about your devices from several tools into one record per device, scores each device against your standards, and tests possible root causes against the evidence.
Which tools can it connect to?
It reads CSV exports, REST APIs that return JSON, files or object storage, SQL databases (read-only), Kafka batches and Google Sheets shared by link. The sample workspace models a directory, a CMDB, an endpoint-management tool, a digital-experience tool and a service desk as CSV exports. Only some connectors are verified; each one lists its limits. Databricks and Salesforce are planned but not available yet.
Do I need to write code or SQL?
No. Standards, checks, KPIs and goals are configured without code, and pipelines are built on a canvas. If you want it, the same pipeline is a readable Python script. There is no SQL authoring.
Where does my data go?
Source rows are held in your session while you work. What is saved is your configuration, run history and aggregate trend snapshots. Data is only sent to an AI provider if you add your own key and use the assistant.
Do I need an AI key?
No. The assistant answers from built-in rules without one. Add a key later for richer, free-form conversations.
How many devices can it handle?
The sample fleet has 420 devices. Pipelines run in memory on the machine that hosts FAAR, so the practical limit is that machine's memory; tables and charts adapt to stay readable.
Can I change the standards and the root-cause library?
Yes, all of it, from the catalog administration page - no coding and no restart.
See it on a sample fleet in minutes
Create a workspace, load the sample fleet or add your own file, and watch 420 devices come together into one record each.