Flushnet Plans & Pricing

One platform, two usage balances

Flushnet Router / AI Gateway does not create a third wallet: it routes work to the selected AI models, tools, and runtimes. External GPU jobs, managed infrastructure, and consulting can still be billed separately.

AI usage: Verified free models use 0 AI Credits. BYOK provider inference uses 0 Flushnet AI Credits and is billed directly by the user's provider. Managed paid AI usage is charged separately through AI Credits based on the selected model/service. flushnet/auto adds no separate surcharge.

Compare plans

Model availability and variable-cost AI services depend on the selected provider/model. A provider such as OpenRouter may expose both free and paid models, so charging is determined per executed model/request rather than by provider name. BYOK provider inference uses the user's own provider account and 0 Flushnet AI Credits. Command recharges add execution allowance but do not change plan-only feature access. Pro may connect and control multiple eligible native, Android, browser and sandbox nodes; Power adds the paid parallel-agent entitlement and Connect priority. Parallel execution is used only by workflows that support it. App Builder and Agent Builder labels describe the included workflow tier; generation, provider inference and Compute work may still use their separately metered credits, and Power advanced behavior is supplied by Power-only orchestration/parallel entitlements rather than a separate editor.

Current Flushnet AI capabilities

This feature catalog reflects the current production Flushnet AI surface. Plan entitlement, provider availability and usage billing still apply; the table describes access and metering without creating a new wallet or surcharge.

The authoritative access decision remains IAM/RBAC. This catalog is shared with the Commercial & Authorization Control Center so the public feature description and admin view do not drift independently.

AI Runner Multi-Host Compute & Automation

Pricing depends on the number of hosts, execution volume, orchestration requirements, managed infrastructure and support level. Request a quote through the online support form.

Turn a pool of connected computers and devices into one coordinated execution layer for AI-driven automation. This specialized AI Runner package is designed for teams that need workflows to run across multiple machines, operating systems, compute resources and edge devices instead of being limited to one computer.

• Multi-platform testing — build once and coordinate validation across Windows, macOS, Linux, browsers and Android.
• Distributed workflows — assign different workflow stages to different hosts, such as build on Windows, services on Linux, validation on macOS, mobile testing on Android and browser automation in parallel.
• Distributed compute & machine pools — detect available CPU, GPU, RAM and storage, select suitable nodes, distribute workloads, retry failed jobs and aggregate results.
• ETL & data pipelines — extract on one node, transform on another, validate elsewhere and load results into databases, data platforms or warehouses.
• Fleet software operations — coordinate software updates, deployments, service restarts, health checks and rollback workflows across multiple machines.
• Monitoring & troubleshooting — inspect logs, applications, services, network state and resource usage across connected hosts from one coordinated workflow.
• Parallel execution — run AI inference, data processing, builds, validation and browser/mobile tasks concurrently on different nodes.
• IoT & edge automation — include Android/edge nodes for sensors, BLE devices, telemetry, mobile applications and field automation.
• Heterogeneous infrastructure — coordinate Windows, macOS, Linux, Android, browsers, remote compute and supported IoT/edge environments in one workflow.

Commercial model: this is a specialized AI Runner multi-host compute and automation package for distributed execution across connected AI Runner nodes. It is not Flushnet Compute: AI Runner Multi-Host coordinates your connected machines; Flushnet Compute is the separate hosted task/build execution service. Runner Commands meter execution activity. Verified free Flushnet AI routes remain 0 AI Credits; BYOK provider inference is billed directly by the user's provider; Flushnet-hosted or managed tasks consume Flushnet Credits only when that specific task/service is defined as chargeable. Dedicated infrastructure, managed node pools, custom orchestration, onboarding and enterprise deployment can be quoted separately.

Flushnet Compute — Hosted Tasks, Builds & Managed Workflows

Flushnet Compute is the hosted execution and outcome-delivery service. Submit jobs and build workflows that run through Flushnet-managed web, automation, processing and optional external compute backends. It is separate from AI Runner Multi-Host Compute & Automation, which coordinates work across your connected AI Runner machines and devices.

• Simple & managed tasks — data extraction, browser automation, AI batch processing and document processing.
• Build & Deliver projects — submit larger outcomes, track plans, jobs, milestones, progress and deliverables in one workspace.
• App & Mobile Builder — generate, iterate, validate, preview, export and deploy supported web/mobile application projects.
• Presentation Builder — create presentations from a prompt and produce supported Office deliverables.
• Audio transcription workflows — browser/device, Flushnet-hosted and explicitly authorized managed STT routes.
• RAG / Knowledge Base processing — document extraction, cleanup, chunking, embeddings, indexing and retrieval workflows.
• ETL / ELT services — extract, transform, enrich and load data into files, APIs, databases and data warehouses.
• Managed GPU / AI compute requests — request accelerated inference, Python/CUDA, training, embeddings and batch workloads when additional compute is required.
• Model training workflows — dataset-to-training execution using supported user-provided or managed provider/compute routes.
• Job queue & deliverables — track execution status, outputs, files and completed results from one hosted workspace.

Commercial model: Flushnet Compute is usage-based rather than a separate monthly multi-host subscription. Compute tasks consume Compute Credits only when the task/service defines a Compute charge. Free AI model routes remain 0 AI Credits; BYOK provider inference uses 0 Flushnet AI Credits and is billed by the user's provider; managed paid AI inference uses AI Credits after the applicable authorization. External GPU capacity and other separately quoted infrastructure are not started without confirmation.

Audio Transcription / Speech-to-Text

Upload audio in Flushnet Compute and process it as a queued transcription job with editable transcript history and downloads.

• Workflow billing: 10 Compute Credits + 1 per started minute, with a file-size floor.
• AI Accuracy: managed STT through Flushnet Router; actual provider inference uses separate AI Credits after explicit confirmation.
• Flushnet Hosted STT: faster-whisper fallback on Flushnet hosting; 0 AI Credits, intended as an economy/private option rather than the highest-accuracy route.
• Downloads: TXT and JSON for completed transcripts; SRT/VTT when timed segments are available.
• Current direct-upload limit: up to 120 MB. Managed high-accuracy jobs automatically use the Flushnet FFmpeg media worker to extract video audio and split sources above the direct STT limit into safe chunks before Router transcription; timestamps are merged afterward. Larger media can use the GPU Compute request path.
• Uploaded source audio remains owner-scoped in the Compute workspace until the user deletes the transcription.

Open Audio Transcription

RAG / Knowledge Base & Embeddings

Turn documents, notes, policies, manuals and remote folders into an owner-isolated AI knowledge base that can be reused by Flushnet AI, Agent Builder, App Builder and API integrations.

• One workflow: extraction → cleanup → chunking → semantic embeddings → vector/index build → retrieval.
• Bring Your Own Storage: external documents stay in the user's own WebDAV/Nextcloud or other connected storage account through Flushnet Connect. Flushnet does not resell shared-hosting disk as customer cloud storage.
• Temporary ingestion: manual uploads and remotely fetched files are deleted after conversion/indexing; the Knowledge Base retains extracted/indexed data required for retrieval.
• Free embedding route: verified zero-cost embedding inference uses 0 AI Credits. RAG processing/indexing still uses Compute credits.
• Managed embedding route: Flushnet pays the provider and normal plan-aware AI Credit accounting applies.
• BYOK: embedding provider inference uses 0 Flushnet AI Credits; the external provider bills the user's own account.
• Free-only mode: never falls back to a paid embedding model.
• Auto mode: tries suitable verified-free semantic embeddings first, then a managed route when necessary.
• Private baseline: hashed vectors are available as a zero-external-cost fallback, but are not marketed as equivalent to a learned semantic embedding model.
• Embedding model/dimension changes require a controlled re-index so incompatible vectors are never mixed.

Plan indexed-corpus limits: Free 50 documents / 50 MB / 10 MB per file; Pro 1,000 documents / 2 GB / 25 MB per file with connected-storage sync up to 250 files per sync; Power 10,000 documents / 20 GB / 50 MB per file with connected-storage sync up to 1,000 files per sync. These are indexing/processing allowances, not hosted file-storage quotas. RAG Compute charges are separate from Runner Commands and AI-provider Credits.

Open RAG / Knowledge Base · Open Flushnet Compute · Open Agent Builder

Command Credits — recharge anytime

Need more commands before your normal plan allowance becomes available again? Buy a one-time command top-up. Purchased credits stay on your account across daily/monthly resets and are consumed only after the included plan allowance is exhausted.

• 1,000 commands — €3.00 Recharge
• 5,000 commands — €12.00 Recharge
• 20,000 commands — €35.00 Recharge

Credits are account-scoped and shared across Access Code, Connect v2, MCP, browser, desktop, and Android execution paths. A recharge adds command allowance only; it does not unlock paid-plan native features.

View plan usage and credit balance

AI-MAX — private local AI

AI-MAX helps your Macs, Windows PCs, and local AI Runner nodes work together as one private AI engine for larger language models—without cloud inference.

• Run larger models across your own machines
• Keep data on your private network
• Use local and offline-capable workflows

View AI-MAX installer · Ask about AI-MAX

Chronos-DX Models — inquiries

Ask about the Chronos-DX model family for text, image, and video use cases, evaluation, integration, deployment, and business requirements.

• Chronos-DX Text — language, reasoning, coding, technical, and structured workloads
• Chronos-DX Image — image generation and visual workflows
• Chronos-DX Video — video generation and multimodal workflows
• Evaluation, integration, deployment, licensing, and custom project inquiries
• Availability and deployment options depend on the model and use case

Ask about Chronos-DX models

Agent Builder & Background Agents

Agent Builder is part of the Flushnet platform plans. Build web agents, deterministic agents, AI-assisted agents, and hybrid workflows through AI Runner, Workflow Manager, Aicron, AI Gateway, and local AI Runner nodes.

• Free: limited Agent Builder access for evaluation and basic workflows
• Pro: Agent Builder plus standard workflows and scheduled automation
• Power: advanced background/scheduled agents, parallel-agent workflows, advanced multi-node orchestration, and Connect priority entitlement
• BYOK external AI providers are supported where available; provider inference is billed directly by that provider to the user and uses 0 Flushnet AI Credits
• Dedicated managed runtimes, enterprise policies, and custom private deployments may still require activation or a custom package

Open Agent Builder · View Public Demos · Ask about managed Agent Runtime

Model Training & Inference – activation on request

Request access to Flushnet AI Model Builder for dataset preparation, model training workflows, local AI Runner training, private compute setup, and inference routing.

Bring your own compute.

Flushnet can run AI workloads through external GPU jobs when needed, or help you create a private local compute network using your existing laptops, desktops, workstations, GPU machines, and connected AI Runner nodes.

Instead of sending every training or inference job to paid external GPU services, you can use your own machines whenever possible and reserve external GPU compute only for heavier workloads. This can lower costs, improve data control, and make better use of hardware you already own.

• Dataset upload, validation, cleaning, deduplication, and ETL / Data Prep
• Dataset Studio and CSV to LLM JSONL conversion
• Training workflows through external GPU jobs or local AI Runner nodes
• Local AI Runner training on connected Mac, Windows, or Linux machines
• Use existing laptops, desktops, workstations, and GPU machines as private compute nodes
• Optional private multi-node training and inference infrastructure
• LoRA / adapter-style fine-tuning workflows
• Inference routing through AI Gateway after model registration
• External GPU job requests for larger workloads when local compute is not enough
• Managed setup available for teams that want help configuring local or hybrid infrastructure

Training, inference, GPU usage, hosted provider costs, external compute jobs, private infrastructure setup, and managed support may be billed separately from the main AI Runner plan.

Request Model Training & Inference activation

Business & Technical Consulting – on request

Need help building your AI project, automation system, data workflow, local compute setup, or just getting started with basic AI Runner usage? Request business and technical consulting for guidance on your own project.

Flushnet can help you review your idea, choose the right execution model, understand what is technically possible, estimate complexity, compare local vs external compute options, and create a practical implementation roadmap.

• Basic AI Runner usage help and onboarding
• AI project feasibility review
• Business workflow and automation analysis
• Technical architecture advice
• AI agents, Workflow Manager, Aicron, and AI Gateway planning
• Local/private infrastructure and compute strategy
• Dataset, model training, and inference planning
• Cost, privacy, and deployment trade-off analysis
• Custom integration planning
• Roadmap, milestones, and implementation checklist

Consulting, discovery sessions, architecture design, implementation planning, onboarding help, and custom advisory work may be billed separately from the main AI Runner plan.

Request Business & Technical Consulting

GPU / AI Jobs

GPU / AI Jobs are additional external compute requests billed separately from the main plan.

• Requested separately
• Charged by usage tier
• LLM, Whisper, Python, CUDA
• Additional external remote resources available on request

Request additional external remote resources

Enterprise / Teams

Custom package for teams, agencies, companies, and managed automation deployments.

• Multiple users and shared team access
• Managed nodes, browser profiles, and workflow orchestration
• Custom usage limits, priority support, and onboarding
• Optional private integrations, approval flows, and dedicated automation setup

Request Enterprise / Teams package