FAQs
General
Section titled “General”What is Fleet?
Section titled “What is Fleet?”Fleet is an AI-powered SOC Analyst developed by Binalyze. It assists security professionals with threat hunting, detection engineering, digital forensics, incident response, malware analysis, and threat intelligence. Fleet provides a conversational interface where you can upload evidence, ask questions, and receive structured analysis results.
Who is Fleet designed for?
Section titled “Who is Fleet designed for?”Fleet is built for SOC analysts, incident responders, threat hunters, detection engineers, and forensic investigators. Anyone who works with security operations and has a Fleet license can use it.
Which models run our work, and on whose account?
Section titled “Which models run our work, and on whose account?”That depends on the billing mode you choose, and it is the same answer as the pricing page gives. On BYOK the models run on your own OpenAI, Anthropic or Google account, under the agreement and data terms you already have with that provider. On the AI Gateway they run through Binalyze’s gateway on Binalyze’s provider accounts, and your plan includes the credits.
How does Fleet differ from a general-purpose AI chatbot?
Section titled “How does Fleet differ from a general-purpose AI chatbot?”Fleet is a specialized security analyst, not a general-purpose chatbot. It runs real DFIR tools, produces verifiable outputs (detection rules that compile, analysis reports with hashes and timestamps, STIX bundles that conform to the standard), and integrates directly with AIR for endpoint operations. When Fleet analyzes a binary, it performs actual static analysis with industry-standard tools. When it generates a YARA rule, it validates that the rule compiles and can scan files with it.
Credits and Sizing
Section titled “Credits and Sizing”How do Fleet Credits actually get consumed?
Section titled “How do Fleet Credits actually get consumed?”A credit measures the work an investigation does through Binalyze’s AI Gateway. Every step (reading the evidence, deciding what to do next, running forensic tools in the sandbox, interpreting their output, writing the report) draws on more than one resource: model tokens, the computing power that runs the tooling and the specialist agents, and the storage and delivery around them. Credits meter that work as a whole, so what a run costs depends on how much evidence gets read, how much tooling runs and how much reasoning it takes, not on a token count. Work you run on your own provider key never touches Fleet Credits at all; your provider bills those tokens directly.
Are Fleet Credits tokens? Is one credit one token?
Section titled “Are Fleet Credits tokens? Is one credit one token?”No. A Fleet Credit is Fleet’s unit of investigation work, not a token. AI model tokens are only one of the resources an investigation consumes. Every run also uses the isolated sandbox it executes in, the computing power behind the forensic tooling and the specialist agents, the storage for evidence and artifacts, and the delivery of the finished report. Credits cover all of that together, which is why a credit cannot be expressed as a number of tokens. Model rates differ too: tokens a model writes cost more than tokens it reads, and a lighter model goes much further than the flagship model. That is why we publish measured credits per complete investigation rather than a token conversion. The average is about 120, and the full table is on the pricing page.
How many credits does a typical investigation use?
Section titled “How many credits does a typical investigation use?”About 120. We ran each of the 20 sample investigations on Fleet’s home screen repeatedly on the AI Gateway’s default model and recorded what every run consumed. The lightest (listing AIR cases) landed near 10 credits. The heaviest (building, converting and testing Sigma rules from a full attack scenario) landed near 239. The complete table, with the observed range for every investigation, is on the pricing page under How far do credits go? (open See all 20 benchmark measurements).
Is there a per-agent or per-task price, report writing versus PST versus PCAP?
Section titled “Is there a per-agent or per-task price, report writing versus PST versus PCAP?”There is no per-agent tariff, and we would rather say so plainly than publish one we do not meter. Credits are not a menu of agent SKUs. They meter investigation work on the AI Gateway, so what a run costs depends on how much evidence gets read and how much reasoning it takes, not on which specialist agent picks it up. That is why we publish complete investigations end to end instead. The closest per-task figures we have are measured: a full encrypted-PCAP analysis at about 100 credits (69 to 126 observed), a Windows event-log triage at about 55, a macro-document verdict at about 111. A finished report is part of those runs, not an extra line.
What does a large mailbox analysis cost?
Section titled “What does a large mailbox analysis cost?”We have not measured one, so there is no number on this page. Mailbox work scales with how much of the mailbox an agent has to read, which varies more than any other evidence type we have benchmarked. If mailbox volume is what decides your plan size, write to [email protected] and we will size it against your real data rather than a published average.
How should we size a Professional plan?
Section titled “How should we size a Professional plan?”Start from the measured 120-credit average. Professional includes 10,000 credits a month on the AI Gateway, which is roughly 80 complete investigations, or about 4 every working day. MSSP/Enterprise includes 50,000 per named analyst, roughly 416 a month. Credit packs cover a busy month without changing the plan, and on BYOK the allowance stops being a ceiling entirely because your provider bills the tokens.
BYOK and Billing Modes
Section titled “BYOK and Billing Modes”What does Bring Your Own Key (BYOK) mean in Fleet?
Section titled “What does Bring Your Own Key (BYOK) mean in Fleet?”BYOK means you connect an API key from a model provider account you already pay for, and Fleet runs every model call on that account. Supported providers are OpenAI, Anthropic and Google, including Anthropic models through Amazon Bedrock and OpenAI models through Azure OpenAI. Your provider bills the tokens at your contracted rates, on your invoice, under your data terms and spend limits. Fleet charges for the platform seat only, which is why a BYOK seat costs less and includes no Fleet Credits. Work on your key never consumes Fleet Credits.
Is BYOK an encryption key? Does it keep our prompts away from the model vendor?
Section titled “Is BYOK an encryption key? Does it keep our prompts away from the model vendor?”No. BYOK is a billing and account choice, not an encryption feature. The prompt and the evidence the agent reads are still sent to the model you selected, and that provider processes them under its own terms. What changes is whose account the call runs on: your contract, your regional endpoint and your data terms apply instead of Binalyze’s. If your reviewers are asking about customer-managed encryption keys for data at rest, that is a separate question. Send it to [email protected] and you get a written answer against your contract.
What exactly do we paste into Fleet?
Section titled “What exactly do we paste into Fleet?”An API key, never a login or password. You create it in the provider’s developer console: the OpenAI Platform, the Anthropic Console, or Google AI Studio. For Amazon Bedrock you enter an AWS access key ID and secret access key with Bedrock invoke permissions, plus the region. For Azure OpenAI you enter the resource key and resource name. A ChatGPT or Claude chat subscription is not an API account; you need the provider’s API or cloud account with billing enabled. An administrator of that account usually creates the key. We recommend a dedicated key for Fleet with its own spend limit, so you can rotate or revoke it without touching anything else. Fleet verifies the key against the provider before saving it.
How is our key stored and used?
Section titled “How is our key stored and used?”The key is saved write-only to your encrypted secrets store and is never shown again. The agent sandbox only ever sees a masked placeholder. The real value is injected at Fleet’s egress edge on the outbound call to your provider, and that call goes straight to the provider’s endpoint, your AWS region, or your Azure resource. The moment a provider is connected, Binalyze’s own model keys are switched off for your account, and every turn, including background tasks, runs on your key. Fleet’s engine and agents do not change between the two billing modes.
On BYOK, how do our provider’s tokens map to Fleet Credits?
Section titled “On BYOK, how do our provider’s tokens map to Fleet Credits?”They do not. On BYOK the model tokens are billed by your provider, at your provider’s prices, and nothing is converted into credits. Fleet never marks up a token. The rest of the platform, from the sandbox to the forensic tooling, is covered by your seat.
Does every AI Gateway plan include 50,000 credits?
Section titled “Does every AI Gateway plan include 50,000 credits?”No. Analyst includes 3,000 credits a month, Professional 10,000, and MSSP/Enterprise 50,000 per named analyst. Included credits renew on your billing date. Credit packs cover busier months and never expire. The 50,000 figure is the MSSP/Enterprise allowance only.
Can we switch between BYOK and the AI Gateway?
Section titled “Can we switch between BYOK and the AI Gateway?”Yes, at any time, with no migration. Switching billing mode changes the seat price and whether Fleet Credits are included; nothing else about your plan, your workspaces or your investigations changes. Connect a key and your account runs on it; remove it and your account returns to the AI Gateway.
Plans and Quotes
Section titled “Plans and Quotes”What does “custom” mean on MSSP/Enterprise?
Section titled “What does “custom” mean on MSSP/Enterprise?”The seat price is published, not custom: $995 per analyst on your own key, $1,495 with the AI Gateway and 50,000 credits per analyst included. What is custom is the envelope around the seat: report and agent allowances, concurrency, connector scope, volume seat pricing, workspace retention, and invoicing and contract terms. There is no hidden price list behind Talk to us.
How do we get a quote?
Section titled “How do we get a quote?”Email [email protected] with your analyst count, whether you want to run on your own model keys or on the AI Gateway, and which connectors you need. You get a written quote built on the published seat prices and the allowances you asked for. Analyst and Professional need no quote at all: start the 7-day identity-verified trial and subscribe in the app.
Evidence and Uploads
Section titled “Evidence and Uploads”What is the upload limit?
Section titled “What is the upload limit?”A verified seat attaches up to 50 files in one prompt, up to 4 GiB per file and 8 GiB across the prompt. A trial seat that has not finished identity verification gets a smaller envelope (10 files, 100 MB per file, 500 MB per prompt) until it does, which takes about 90 seconds. Uploading is not the only way evidence arrives: the workstation has a real terminal, so agents also fetch, extract, carve and generate files inside the workspace.
Can I analyze a 10 to 20 GB PST?
Section titled “Can I analyze a 10 to 20 GB PST?”Not as a single upload today. The ceiling is 4 GiB per file, so a mailbox above that has to arrive in pieces. We are not publishing a date for a larger single-file limit, and we would rather leave this answer uncomfortable than promise one. If the bigger envelope decides the deal for you, tell us at [email protected] so it is counted.
So what do we do today when the evidence is larger than the cap?
Section titled “So what do we do today when the evidence is larger than the cap?”Reduce it before it travels, or bring it in parts. Export the folders, the custodians or the date range that the case actually turns on into smaller PSTs and attach those. Attach the artifact instead of the container: the event-log set, the registry hives, the carved directory, a single process dump rather than the whole memory image. An archive split across parts can be reassembled and unpacked in the workstation terminal. And with the AIR connector, query what AIR already parsed from that endpoint through the investigation data lake instead of moving the raw evidence at all.
Security and Privacy
Section titled “Security and Privacy”Is my data sent to third parties?
Section titled “Is my data sent to third parties?”No customer evidence data is stored or shared externally. All analysis runs in a secure, isolated environment managed by Binalyze. On the AI Gateway, AI communication is routed through Binalyze-managed secure endpoints. On BYOK, model calls go straight to your own OpenAI, Anthropic or Google account and do not pass through Binalyze’s AI Gateway. No customer data is used to train AI models.
Are my API keys safe?
Section titled “Are my API keys safe?”Yes. API keys and credentials configured for AIR integration are encrypted at rest and in transit. They are handled by a secure gateway that is architecturally separate from the AI agent. The AI agent never sees, handles, or has access to your actual API keys. See Security for full details.
Can Fleet access my production systems?
Section titled “Can Fleet access my production systems?”Fleet operates in an isolated environment and cannot directly access your network, endpoints, or internal systems. It can interact with AIR endpoints only through the authenticated API integration you configure. All endpoint operations go through AIR’s standard API with your configured permissions.
How is one client’s evidence kept away from another’s?
Section titled “How is one client’s evidence kept away from another’s?”Every investigation runs in its own isolated workspace with its own sandbox. Evidence, artifacts and reports never cross workspace or tenant boundaries, agent egress is limited to a policy-defined allowlist rather than the open internet, and retention is yours to set: export or destroy on your schedule. The Security and Trust page has the detail.
Our MSA restricts where client data may be processed. Who answers that?
Section titled “Our MSA restricts where client data may be processed. Who answers that?”We do, in writing, to your security and legal teams. Send the questionnaire or the clause to [email protected] and you get a specific written answer rather than a marketing paragraph. We would rather answer a residency or data-processing clause precisely, against your contract, than post something general here and have you discover the gap in review.
Capabilities
Section titled “Capabilities”What file types can I upload to Fleet?
Section titled “What file types can I upload to Fleet?”Fleet accepts any file type. Common formats include:
- Executables: PE (.exe, .dll), ELF, .NET assemblies, shellcode
- Documents: PDF, Word (.doc, .docx, .docm), Excel, PowerPoint, RTF, OLE
- Archives: ZIP, 7z, RAR, TAR, GZ (including password-protected)
- Network captures: PCAP, PCAPNG
- Event logs: EVTX, JSON, JSONL
- Detection rules: YARA (.yar), Sigma (.yml), osquery (.sql), Suricata (.rules)
- Images: PNG, JPG, GIF, BMP
- Text: TXT, CSV, MD, JSON, XML, HTML
- Disk images: RAW, E01, VMDK, VHD/VHDX
- Memory dumps: Raw memory dump files
What detection rule formats does Fleet support?
Section titled “What detection rule formats does Fleet support?”Fleet supports four detection rule formats:
| Format | Capabilities |
|---|---|
| YARA | Generate, validate, fix, scan files |
| Sigma | Generate, validate, convert to Splunk SPL and Microsoft Sentinel KQL, test against EVTX and JSONL logs |
| osquery | Generate, validate syntax and column references |
| Suricata | Validate rule syntax |
Can Fleet deploy detection rules to my endpoints?
Section titled “Can Fleet deploy detection rules to my endpoints?”Yes, when connected to AIR through the API integration. Fleet can deploy YARA, Sigma, and osquery rules to managed endpoints via AIR’s triage feature. See AIR Integration for details.
What operating systems can Fleet analyze artifacts from?
Section titled “What operating systems can Fleet analyze artifacts from?”Fleet can analyze artifacts from Windows, macOS, Linux, ChromeOS, and ESXi. The analysis environment itself runs Linux, but it includes tools capable of analyzing artifacts from all these platforms (PE analysis for Windows, plist parsing for macOS, ELF analysis for Linux, etc.).
Can Fleet analyze password-protected archives?
Section titled “Can Fleet analyze password-protected archives?”Yes. Fleet automatically tries common malware analysis passwords (infected, malware, virus) when encountering password-protected archives. You can also provide the password in your prompt, or Fleet will ask you for it if the common passwords do not work.
What threat intelligence outputs does Fleet produce?
Section titled “What threat intelligence outputs does Fleet produce?”Fleet produces threat intelligence in multiple formats:
- Markdown reports — human-readable reports with observables organized by type, enrichment results, and risk scores
- Text reports — flat lists sorted by risk score for quick triage
- STIX 2.1 bundles — machine-readable structured intelligence with full enrichment data, importable into SIEM and TIP platforms
- YARA rules — detection rules generated directly from enriched observables
AIR Integration
Section titled “AIR Integration”How do I connect Fleet to my AIR deployment?
Section titled “How do I connect Fleet to my AIR deployment?”In Fleet’s settings, locate the AIR integration section and provide your AIR API key. Fleet validates the connection and confirms access. Once configured, Fleet can perform endpoint management, evidence acquisition, triage, interACT, and investigation operations.
What AIR operations can Fleet perform?
Section titled “What AIR operations can Fleet perform?”Fleet can perform the following operations through AIR:
- Endpoint management — list, search, isolate, remove isolation, tag, reboot
- Case management — create, update, close cases
- Evidence acquisition — trigger collection using built-in or custom acquisition profiles
- Triage — deploy YARA, Sigma, and osquery rules to endpoints
- interACT — execute remote commands, transfer files
- Investigation — browse evidence, view DRONE findings, access Investigation Hub data
Does Fleet respect AIR’s role-based access control?
Section titled “Does Fleet respect AIR’s role-based access control?”Yes. Fleet inherits the user’s AIR session authentication and respects role-based access control. Users can only perform AIR operations that their assigned role permits.
Can I point Fleet at an AIR case and ask questions about the whole case?
Section titled “Can I point Fleet at an AIR case and ask questions about the whole case?”Yes, for everything AIR holds about that case. With the AIR connector configured, agents work it the way an analyst would: resolve the case by id or name, read its notes, linked assets, tasks, assignments and audit trail, pull the asset inventory with each asset’s acquisition and triage history, and query the Investigation Hub data lake with read-only SQL across its findings, timeline and full-text-search tables. Actions that reach live endpoints, like starting a triage or an acquisition, ask for your approval before they run.
Does the raw evidence come with it?
Section titled “Does the raw evidence come with it?”No, and that distinction is the one worth getting right before a first call. The connector reaches what AIR has already collected and parsed. A disk image, a memory image, a mailbox or a packet capture becomes analyzable once it is in that investigation’s isolated workstation: you attach it, or an agent downloads or acquires it there first. Disk and memory images, mailboxes, packet captures, log archives, binaries and anything from outside AIR have to land in the workspace before file-level analysis starts.
Can I use Fleet offline?
Section titled “Can I use Fleet offline?”No. Fleet requires an active internet connection and is available only in online mode.
What AIR version is required?
Section titled “What AIR version is required?”Fleet requires AIR version 4.41 or later.
Do I need a separate license for Fleet?
Section titled “Do I need a separate license for Fleet?”Fleet is licensed separately. Contact Binalyze sales or your account representative for licensing details.
How do I access Fleet?
Section titled “How do I access Fleet?”Fleet is a standalone web application. Log in at your organization’s Fleet URL with your credentials. No additional software installation is required.
Can multiple users use Fleet simultaneously?
Section titled “Can multiple users use Fleet simultaneously?”Yes. Each user gets their own isolated Fleet session. Multiple users can use Fleet concurrently without interfering with each other’s work.