AVA
A desktop sales intelligence app for German-speaking B2B teams. It researches target accounts in the Handelsregister and Bundesanzeiger, watches LinkedIn for openings, and flags the moment a prospect is worth calling.
What is AVA?
AVA is a sales intelligence application for B2B teams selling into German-speaking markets, and the first thing to understand is that it is a desktop program for Windows and macOS, not a browser SaaS. The question it sets out to answer is a practical one: when should you contact whom. It works from nine declared sources. The commercial register supplies master data, legal form, authorised representatives and capital; the federal gazette supplies filed accounts, balance sheets and profit and loss figures for companies that must publish them; company websites, LinkedIn, existing CRM records, Excel or CSV imports, map data, targeted web searches and the public register stock make up the rest. Every result carries what the vendor calls a Fundstelle - the specific place the fact was found - with a link back to the original. Three functions sit on top of that. Research enriches a list of accounts you already have, imported from a spreadsheet or pulled from HubSpot, Salesforce, Microsoft Dynamics 365 or Notion. Discovery works the other way: you supply your own website plus the websites of your best customers, the tool infers an ideal customer profile from them in about two minutes, and then hunts your region for companies that match it. Nothing is imported automatically - you approve each hit. Monitoring runs in the background and pushes alerts into an inbox, optionally onto Telegram, when something happens: a managing director changes, a company files accounts showing growth, a contact posts something worth reacting to. The distinguishing choice is where the AI runs. You can point it at a local Ollama model, bring your own key for OpenAI, Anthropic, Google, Mistral, DeepSeek, xAI or Qwen, or run your own model server. A quality tier from S to C is shown per company so you know whether a profile was written by a frontier model or a small local one, and a daily spend cap is available for cloud models. The LinkedIn module reads only. It never likes, comments or messages, and it requires explicit timestamped consent. The vendor is unusually frank that using it carries some risk to the account.
What it does
- Research a target company automatically across German registers, its website and LinkedIn
- Discover new prospects in a chosen region that match an automatically derived ideal customer profile
- Enrich an Excel list or a connected CRM without manual lookups
- Get alerted on director changes, insolvencies and revenue jumps
- Rank a pipeline by AI scoring and by how strong each signal is
- Read a LinkedIn feed for contact openings without posting or messaging
- Trace every result back to the register entry or page it came from
When to use AVA / When not to
A quick filter to help you decide if AVA is the right fit.
When to use AVA
- Field sales reps covering a German region who need a reason to call before they knock on the door
- SDR and BDR teams that want their list ranked by signal strength instead of by import order
- Sales managers who need to see which accounts in the existing book have just moved
- Companies selling into German manufacturing, where filed accounts and register entries are richest
- Small teams and solo sellers who want a permanent free tier rather than an enterprise contract
When not to use AVA
- Anyone selling outside the DACH region, since the core sources are German public registers
- Teams that need email sequencing or outreach automation, which the vendor states is not included
- Buyers looking for a verified phone and email database, explicitly outside the product's scope
- Marketers who want website visitor tracking, a feature the vendor says it deliberately omits
- Linux users and anyone needing a native mobile app, as builds ship only for Windows and macOS
How to use AVA
A typical end-to-end flow, from setup to results.
- Download the installer for Windows or macOS from the vendor's public release page
- On macOS, launch once so the quarantine check runs, then quit and reopen before going further
- Enter your own website address and the addresses of your best existing customers
- Review the ideal customer profile the app derives from them and correct it if needed
- Choose how the AI should run: a local model, your own cloud API key, or your own model server
- Set a daily spend cap if you opted for a cloud provider
- Import your company list from Excel or CSV, or connect HubSpot, Salesforce, Dynamics 365 or Notion
- Let the background research run; expect roughly thirty seconds to two minutes per company
- Switch on the regional radar and approve or ignore the new matches it proposes
- Enable the LinkedIn module and Telegram alerts only if you accept the caveats attached to each
Pros & Cons
Pros
- Unusual depth on German companies, including filed accounts and management history
- You choose where the AI runs, which very few tools in this category offer
- Every fact is traceable to the register entry or page it came from
- A permanent free tier of twenty-five companies with no card and no time limit
- Published prices that sit well below the enterprise contracts typical of this market
- LinkedIn is read only, with no automated liking, commenting or messaging
- The code repository is public and releases ship almost daily
Cons
- The vendor's own comparison pages price the product at zero and call it open source, while the pricing grid, the structured data and the terms all describe paid monthly plans
- The repository carries no licence file, and its README calls the project internal with external contributions not envisaged, so the open source claim is not backed up
- The terms say no personal data reaches the vendor's servers, but the privacy policy and FAQ describe a central database in Frankfurt
- Usefulness falls away sharply outside German-speaking markets, since the sources are German
- Several components are still described as in progress, including automatic deletion after 180 days
- No data processing agreement, no formal subprocessor list and no documented training opt-out
- No Linux build and no native mobile app; mobile access runs through Telegram
Pricing & Plans
A permanent free plan is available: it covers twenty-five companies for the lifetime of the account, includes all functions, requires no payment card and carries no time limit. The lowest paid entry point is the Starter plan at EUR 49 per month, quoted before the applicable 19% German VAT, which raises the monthly allowance to five hundred companies. Pro follows at EUR 149 per month on the same terms for two thousand companies, and an Enterprise tier is offered on request. Billing runs through Stripe and subscriptions renew monthly unless cancelled. One cost sits outside the subscription: if a cloud AI model is chosen instead of the free local one, the provider bills the customer directly for usage.
- EUR 0 - 25 companies for the lifetime of the account
- all functions
- running locally on your device
- EUR 49 per month plus 19% VAT - 500 companies per month
- all functions
- priority support
- EUR 149 per month plus 19% VAT - 2
- 000 companies per month
- all functions
- early access to beta features
- price on request - unlimited companies
- dedicated support
- custom integrations and SLAs
Data, GDPR & hosting
A consolidated view of how AVA handles your data.
GDPR overview
GDPR references are concrete. The controller is named as eproX GmbH in Herford with a dedicated address at datenschutz@eprox.de. The policy, dated September 2026, lists the full set of rights: access, rectification, erasure, restriction, portability, objection, complaint to a supervisory authority, and withdrawal of consent with future effect. It states that no analytics or tracking services are used for advertising purposes, and that only technically necessary cookies may be set. No Article 27 representative is named, which is consistent with a German establishment, and no data protection officer is designated. The important caveat is the vendor's own: asked directly whether the tool is GDPR compliant, the FAQ answers that this depends on the operating model and the customer's configuration, not on the product alone.
Who owns the data?
The customer's own organisation stays the data controller for whatever it processes; the vendor positions itself as a supplier of provenance, deletion mechanisms and templates rather than a legal basis. Data is split three ways: encrypted account data held per account on the machine, public company research sent to the vendor's Frankfurt substrate, and minimal quota telemetry. The licence granted is simple and non-transferable for the contract term, with resale, rental and sublicensing forbidden. Note a genuine conflict in the vendor's own documents: the terms state that no personal data reaches the vendor's servers, while the privacy policy and FAQ describe a central database. German law governs, with jurisdiction at the vendor's registered seat.
Reuse rights
Nothing in the published documents grants the customer a general right to redistribute what the tool retrieves. The material itself comes from public sources - the commercial register, the federal gazette, company websites and LinkedIn - and the vendor disclaims any warranty as to its completeness, accuracy or currency, so onward use is at the customer's risk. Results can be written back into a connected CRM, which is the reuse path the product actively supports. Where a cloud model is chosen, website text, register extracts and, if the module is switched on, LinkedIn post text may be passed to that provider under the customer's own contract with it. LinkedIn reading requires explicit timestamped consent, and the vendor is clear that responsibility for consent and employer rules stays with the user.
Data retention & training
Hosting summary
Hosting is split. The substrate that handles public company research runs in Frankfurt, inside the EU, and the vendor describes it as the destination for company and contact research together with its provenance. Encrypted account data stays on the user's own machine, and quota counting sends only a minimal company identifier and a timestamp. Organisations that want tighter control can ask for a self-hosted Docker substrate covering database, bus, identity and gateway, optionally with their own model server, which keeps flows inside their environment; the vendor describes this as configured jointly rather than self-service. Two caveats. The marketing website itself is hosted with Vercel Inc. and its resolved address geolocates to the United States, which is a different flow from the product substrate. And if a cloud AI model is selected, processing location then depends on whichever provider the customer has contracted with.
Things to keep in mind
Risks and trade-offs to weigh before adopting AVA.
- Reading a LinkedIn feed with automation can get the account temporarily restricted; the vendor says so plainly, and a separate account is a common workaround
- GDPR responsibility stays with your organisation - the tool supplies provenance and deletion mechanisms, not a legal basis for processing prospect data
- Marketing pages contradict the pricing grid and the terms on price and licensing; confirm what you are buying before you sign
- The vendor disclaims any warranty on the completeness or accuracy of the public data it retrieves, so a bad register read is your problem, not theirs
- Cloud model usage is billed by the provider outside the subscription and can grow quietly; set the daily spend cap early
- Subscriptions renew automatically each month unless cancelled before the period ends
- Automated signals invite reflexive calling - a director change is a fact, not a qualified opportunity, and treating the alert as the research is how a pipeline fills with noise
Setup & Integrations
Technical difficulty
Low for the standard path. The vendor promises a double-click application with no dependencies to install and no command line, and the profile setup takes roughly two minutes from a few website addresses. macOS users must launch once, quit and reopen so the quarantine check completes, after which updates arrive automatically. Two options raise the bar: running a local model means running Ollama on your own machine, and the self-hosted Docker substrate is explicitly described as configured together with the vendor rather than installed alone.
Deployment
Integrations
Supported languages
Behind AVA
Social
Resources
All the official URLs gathered for verification and reference.
Alternatives
Tools that compete with or complement AVA.
Frequently asked questions
Is AVA a web application?
Where does the company data come from?
Do I have to send my data to a cloud AI provider?
Is there a free trial?
What does it actually cost once I outgrow the free plan?
Will it post on LinkedIn for me?
Is the software open source?
Does it work for selling outside Germany?
How long does research take?
Who is behind AVA?
Should you pick AVA?
AVA is a narrow tool that is very good at its narrow thing. If your accounts are German companies, the combination of filed accounts, register history and background monitoring gives you something the big American platforms genuinely do not: a reason to pick up the phone this week rather than a longer list of contacts. The freedom to run the AI locally, through your own cloud key or on your own server is the real differentiator, and the per-company quality tiers are an honest touch that few competitors bother with. Two things should temper enthusiasm. The first is scope: outside German-speaking markets the sources that make this product interesting simply are not there. The second is the vendor's own messaging, which contradicts itself in three places. The comparison pages advertise the product as free and open source, while the pricing grid, the structured data on the homepage and the terms all describe paid monthly plans. The repository is public but carries no licence and describes itself as internal. The terms promise that nothing personal reaches the vendor's servers, while the privacy policy and FAQ describe a central database in Frankfurt. None of this makes the software bad, but it does mean a buyer should confirm in writing what they are actually agreeing to. The product is also young. Public releases still carry a 0.x version number, several components are described as in progress, and there is no data processing agreement or subprocessor list yet - which will matter to a compliance team even though the vendor is refreshingly precise about data flows elsewhere. The permanent free tier makes all of that cheap to evaluate, and that is the sensible way in.
- Choosing a selection results in a full page refresh.
- Opens in a new window.