Frequently Asked Questions

Yes, we've made it easy. The interface is a web page: you create cases, upload documents, and ask questions. There's a little bit of setup on site, but we've made the process easy. You'll want one or two 'administrators' with privileged access for things like adding users and configuring backups, but it's all done through the web interface.

Most servers will be the size of a large tower PC. If you need or prefer a rack-mounted server, we can do that also. Our standard tower form factor is quieter and more convenient for most offices.

EdgeDock is accessed through your web browser from the office or over your VPN, just like you'd access a cloud service. Although you're using a web browser, the traffic and data flow stays on your firm's local network.

EdgeDock works on any network, from a simple ISP gateway (e.g. Xfinity, AT&T) to enterprise networks, hardwired Ethernet or WiFi. If you just have a standard ISP router or simple network, we can set it up with a local address using mDNS (like http://edgedock.local:8000) without any network administration. If you have a more advanced network, your network administrators can set it up with a static IP, local DNS, and firewall rules, if they prefer.

Most servers will draw around 100 watts at idle and 600-1500 watts at full load, operating on a standard outlet. Electricity costs are less than $15/month, even at full load. Every system ships with a lithium UPS (battery backup) to protect against data corruption and hardware failure due to power blips.

It's built to operate quietly in your office. When idle, it's nearly silent. Under heavy load, the server will run 35-40 dB (a quiet office conversation). You probably won't notice it above your office's background noise unless sitting right next to it.

EdgeDock can operate fully air-gapped from the internet if you need it to. All AI processing and document storage are done on the server. However, your best results will be with an internet connection, allowing agents to use live web search to retrieve current events and research. We also recommend having internet for off-site, encrypted backups. An internet connection is also needed if you'd like us to maintain and update your system.

It will depend on your selected hardware, your workflows, use cases, and speed expectations. Generally, our lowest-end servers will support up to 10 employees, or 2-4 concurrent users. Better hardware will serve more users and give faster responses. After understanding your needs in our initial conversation, we'll give you our recommendations and options.

EdgeDock is a local-network system by design, for privacy and security. For remote access, you can use your business's VPN to your office network, where EdgeDock will work exactly as it does in the office.

All data between EdgeDock servers, ECS, and backup targets is end-to-end encrypted, always. For human interaction through the browser: We realize that EdgeDock customers will widely vary in their IT capabilities, so it can support a range of implementations. If your network already uses TLS certificates, you can upload your organization's certificate to EdgeDock. If not, EdgeDock includes its own local Certificate Authority, so supports secure HTTPS by loading its certificate onto each individual client computer. For the simplest implementations, EdgeDock allows HTTP access, but with a warning and 'Not secure' will display in the browser.

We use Tailscale and a self-hosted Tailnet to securely access your servers, when you allow it. EdgeDock has a client-side isolation switch that fully disables remote access, and you can establish scheduled windows during which remote maintenance access is allowed. As a backup, ECS uses a physical KVM device to simulate local access, but this can be physically disconnected unless/until needed. If you've contracted with us to do your monthly security patches and upgrades, we'll coordinate a maintenance window during which you'll allow access from EdgeDock. We use secure, encrypted, industry-standard protocols to access your system when you allow it.

Technically, it is possible through system backends, but not necessary or likely and is against the terms in our maintenance contract. Through the same permissions and Case separation that you'll use in normal business, ECS technicians won't be able to see your sensitivie information.

Technology is improving rapidly, but we expect the hardware to have a minimum useful life of 3-5 years. Interestingly, as AI models improve, they are getting more efficient, so your AI may become faster and more capable over time with your same hardware.

The EdgeDock system includes a clustering architecture that lets you add additional servers, intelligently routing requests for best speed and quality. This clustering also supports high-availability failover, protecting your document accessibility and business operations in the event of a system failure.

Possibly. We would need to evaluate your hardware and determine if EdgeDock can support your configuration. Tell us about it in our initial discussion. Custom services and rates may be needed to set it up and test models.

Honest answer: the very top frontier cloud models are still ahead on complex reasoning, nuanced long-form writing, and obscure research. But the open source models that run on EdgeDock today are only about 6-9 months behind the frontier models, and they are completely capable for most tasks. You don't need the bleeding-edge LLM for most work, and you gain the privacy. As open source models improve, you can load them onto the hardware you already own.

EdgeDock runs open-source large language models that are licensed for free commercial use. The specific models will depend on the hardware you select. All EdgeDock servers have a minimum of 32GB VRAM, supporting 30B-class open source models at Q4-Q6 quantizations. Higher-end hardware can allow you to run larger models, run them faster, or serve multiple models in parallel. Our routing and scheduling algorithms calculate the best model and GPU for each request to balance speed and efficiency. EdgeDock is also capable of using OpenAI-compatible cloud endpoints if you choose to route requests to cloud AI providers, despite our recommendation not to. The platform is model-agnostic by design. You are not locked into one model, one vendor, or one generation.

You can, but it's challenging to find the right combination of configuration settings to get them to run well. Also, the inference engine ecosystem usually takes a few months to support new models. That's why Edge Crossing Systems incorporates an 'approved models' list within EdgeDock, allowing you to select models that should run well, with recommended default settings. But in all reality, the models that EdgeDock ships with are fully capable and will serve your needs without upgrade, and you shouldn't need to worry about what model you're running.

Inaccurate results can come primarily from two sources: inaccuracies in the source data (internet or internal documents) and AI model hallucinations. We always need to verify AI's work product, as we would with any junior associate, and competent humans need to provide final review and judgment. To assist with verification, EdgeDock agents always cite their sources, whether your internal documents or websites, with every response.

No, it won't replace those tools but can complement them. But it can take their outputs, as well as your privileged client data, for analysis and document creation. That is, you can research the case law and provide those cases to EdgeDock as documents, where agents can combine it with your privileged client data for analysis and document creation. EdgeDock's 'global' document scope allows you to build a durable library of reference material that is available to all agents for future use.

EdgeDock's outputs are working tools — like a paralegal's notes or a research memo. Humans own the final work product, as if they had prepared it themselves. AI is just a tool, like a word processor or search engine.

We provides the following, with firm, fixed pricing:

• Hardware and Software: Our fully configured servers, with our proprietary EdgeDock software, start around $45,000.
• Remote Maintenance (optional, recommended): ~$7000/year for remote system updates, security patches, EdgeDock upgrades, health monitoring, and technical support. Prefer to maintain it yourself? You can, and the support contract is genuinely optional.

Beyond the system purchase and optional maintenance contract, you should expect only to pay for electricity and web search API costs (3rd party).

It really depends on your plans and usage. For low usage, cloud AI might currently be lower cost. However, it's widely expected that cloud AI costs will significantly increase after this initial 'hook' phase, and local AI can insulate your business. That said, EdgeDock's real value is in privacy and data protection, reducing your liability in the long run.

If you're not using AI yet, you will easily see payback through faster results with less effort, either allowing you to increase your case load (revenue) or decrease costs. Even with very conservative assumptions on time savings, you'll see full payback of your investment within a few months. If you are using cloud AI, EdgeDock can offset your cloud AI costs. But, more importantly, your investment gives you priceless data security and privacy.

AI is great at reducing your time spent particularly on non-billable tasks: background research, document reviews, and preparation. AI can free up your staff's time for more billable hours and taking on more cases.

We don't offer a lease option. For strict privacy and security your data should reside on hardware that you own, with no claims by anyone else. Partial financing or deferred payment structures are possible; we can discuss details.

Because the landscape has changed. The 2026 Heppner ruling, surveillance and data mining, government subpoenas, and increased cybersecurity threats have shown a clear need to protect your sensitive data. AI companies provide confidentiality guarantees on paper but, frankly, we don't trust that unforeseen circumstances won't expose the data. Cloud AI is fine for generic, anonymous research, but you just can't risk uploading sensitive documents or chat prompts anymore. The data you've already uploaded is at risk, but future data doesn't have to be.

Your data — case information, documents, chats — is stored on a dedicated drive, separate from the operating system, and encrypted at rest. Your data never has to leave the server, because the AI inference is done locally. Backups are fully encrypted before leaving the server for off-site cloud storage. Within your firm, data is protected through strict case separation and document permission scoping.

Since it's local to your firewalled network, EdgeDock uses simple usernames and passwords for authentication. For remote maintenance, we use SSH certificates with our self-hosted certificate authority (CA) for scalable and revokable permission management.

Only if granted permission by a case owner. Every case is a sealed workspace with its own permission set. Only other humans with access to a case can see its documents and create agents with visibility to it.

We can tell you exactly what the system does — where data lives, who can touch it, what's encrypted, what gets logged. But whether you're compliant is a question about your policies, your use case, and your agreements (like business associate agreements), and we'd expect you to run that through your own compliance counsel. What we won't do is hand you a sticker that says "compliant." Compliance templates for HIPAA, FERPA, and privilege protection are on our roadmap as a product feature.

Not at all. EdgeDock's agents have a limited, specific set of tools to serve only the expected uses.

EdgeDock comes with a built in document editor built in, so you can directly edit documents from the EdgeDock interface. From the Documents page, you can download individual documents or zip files for email or to import back to your primary document storage.

Yes, EdgeDock employs a combination of keyword search and semantic (vector embedding) search to reliably find documents, both by human users and agents. However, only those humans and agents with permission to the case will find the documents.

Document transfer in and out of EdgeDock is currently manual to ensure that humans keep control over the scoping and visibility. Our Document library makes it easy to drag and drop files into EdgeDock to make them available for agents. When the agent produces documents, it's easy to download them to your file storage.

Yes! EdgeDock can serve as an OpenAI-compatible LLM inference endpoint on your network. Set up an API key in EdgeDock and point your other software to it. EdgeDock will serve the API requests from its resource pool alongside its agent work.

The Document library takes the formats a legal office actually has: PDF, DOCX, TXT, RTF, MD. Each server ships with a dedicated data drive sized for your expected number of users.

Open source local AI is challenging to implement and requires significant technical expertise and experimentation to get it working. We've done that work with the EdgeDock platform, making it easy to integrate AI with your documents and case material.

Hardware and software failures might happen, but we'll be ready for it. EdgeDock supports automated backup of its databases and your documents to either local targets or S3-compatible endpoints for off-site redundancy. All backups are end-to-end encrypted. During initial deployment, we'll help you set up these backups if you wish.

It's unlikely, but possible, and it's an important consideration for business continuity. EdgeDock comes with a one year hardware warranty, but you'll still need to run your business while your server is being repaired. For high-availability (HA) access to your documents and AI services, EdgeDock supports clustering and failover with multiple servers. Depending on your use case, we'll recommend a hardware configuration that gives you best redundancy for managing a hardware failure.

Since you own the hardware and a perpetual license to the software, you would be more insulated than if a cloud provider stopped providing their services. If we ceased operations, you keep the hardware, you keep your data, and the system would continue to run.

You keep everything. All your data, cases, and agents live on your hardware. You retain full access to the system exactly as it was. No data hostage-taking, no feature degradation. You'll lose access to our software upgrades and will need to handle system-level security updates on your own or with another managed service provider.

Privacy and trust. Do you fully trust that your sensitive data uploaded to the internet is protected?