Start in chat
Try the model immediately in the existing chat surface with no new interface to learn.
SELF-HOSTED MODEL PLATFORM
tk_
Self-hosted U.S. models for business software. Specialized by job. Fine-tuned on proprietary continuously collected data, refreshed monthly.
Try tk_ live
Start with the prompt.
Get your API key
Move from prompt testing into production integration without switching products.
Ship against one API
Use the OpenAI-compatible endpoint and route across the full stack as workloads get harder.
PLATFORM
A model company, not a wrapper
Toolkit self-hosts a full business AI stack in the U.S. with continuously refreshed proprietary data pipelines and no external API dependency.
U.S.-hosted inference
Domestic serving and routing discipline for production business workloads.
OpenAI-compatible API
Drop into existing SDKs and tooling without inventing a new client surface.
No external API dependency
The core stack is self-hosted instead of riding on someone else's product decisions.
Specialized model stack
Separate engines for business chat, coding, multimodal analysis, voice, and tool execution.
Continuously collected data
Fresh proprietary domain data is collected continuously and shaped into recurring monthly refreshes.
Production-ready for apps
Built to power downstream business products, agents, and domain workflows.
PRODUCT LADDER
Three buying decisions. One integrated stack.
Publicly, the site stays simple: Base, Pro, and Voice. Internally, the full model system stays specialized.
Toolkit Base
Fast business model for production apps
$0.25 / 1M input · $1.00 / 1M output
General business workflows, research, support, and operational software. Includes cheap tool-task routing under the hood.
Toolkit Pro
Premium code, reasoning, and multimodal work
From $0.40 / 1M input · from $2.50 / 1M output
Software engineering, document-heavy analysis, screenshots, financial reasoning, and harder business decisions.
Toolkit Voice
Real-time voice agents for business workflows
$0.10 / 1M input · $0.50 / 1M output
Phone and conversational surfaces with voice-native operation instead of retrofitted text-first chat.
No weekly lockouts. Rolling limits built for actual work.
MODEL STACK
Specialized models. Routed intelligently.
Customers see simple tiers. Under the hood, the stack separates business chat, engineering, vision, reasoning, voice, and cheap tool execution.
toolkit-base
Core business assistant
$0.25 / 1M input · $1.00 / 1M output
Business apps
toolkit-code
Software engineering and architecture
$0.25 / 1M input · $1.00 / 1M output
Engineering
toolkit-vision
Multimodal document and image analysis
$0.40 / 1M input · $2.50 / 1M output
Vision
toolkit-think
Deep reasoning and financial analysis
$0.40 / 1M input · $4.00 / 1M output
Reasoning
toolkit-voice
Real-time business voice agents
$0.10 / 1M input · $0.50 / 1M output
Voice
toolkit-jr-code
Internal task engine for tool workflows
Routed automatically · No direct billing
Tool execution
ROUTER
One endpoint. Multiple specialized engines.
The inference router decides when work belongs on the base model, the code model, the vision rail, the reasoning rail, or the cheap internal task engine.
User requestToolkit routertoolkit-base
toolkit-code
Software architecture, implementation, debugging, review
toolkit-vision
Documents, screenshots, receipts, multimodal extraction
toolkit-think
Hard business analysis, finance, reasoning, planning
toolkit-jr-code
Linting, transforms, patching, boilerplate, cheap tool ops
toolkit-voice
Real-time voice agents and phone-native flows
DATA MOAT
Continuously collected data. Recurring monthly refreshes.
Freshness belongs to the data operation, not pretend live weights. The moat is collection, shaping, and retraining discipline.
Automotive listings and incentives
Real estate indices and market context
Financial and macro signals
Weather and location-aware demand data
News and demand signals
Framework and release intelligence
collect continuously → validate → shape samples → refresh monthly
18 PRODUCTS · 15 DOMAINS · ONE DEPLOYMENT
The models are the engine. These are the products.
Every vertical runs on the same infrastructure — same models, same API, same billing. One codebase, eighteen faces.
DealerGPT
Real-time vehicle valuations, market timing, and dealer invoice intelligence across every make and model.
Open Site →RealtorGPT
Property valuations, comp analysis, and neighborhood intelligence for buyers, sellers, and agents.
Open Site →BoatGPT
Hull-to-engine valuations for sailboats, cruisers, and yachts. Slip fees, survey costs, total ownership.
Open Site →RVGPT
Motorcycles, UTVs, motorhomes, snowmobiles, jet skis. Seasonal pricing across the $60B rec vehicle market.
Open Site →FlightGPT
Aircraft valuations from Cessnas to Citations. TBO tracking, hangar costs, hourly operating economics.
Open Site →JobGPT
Salary benchmarks, resume optimization, interview prep. Market-rate intelligence across every role.
Open Site →TravelGPT
Route planning, fare tracking, hotel intelligence. Real-time pricing across flights, stays, and experiences.
Open Site →SportGPT
Live odds, player analytics, injury reports, and historical matchup data across every major league.
Open Site →IndexGPT
Market indices, sector rotation, macro signals. Portfolio-grade analysis without the Bloomberg terminal.
Open Site →NewsGPT
Real-time news analysis, source triangulation, and bias detection across global media outlets.
Open Site →MedGPT
Symptom analysis, drug interaction checks, and provider search. Evidence-based health intelligence.
Open Site →PokerGPT
Hand equity, range analysis, ICM calculations. Real-time strategy across cash games and tournaments.
Open Site →A-Aron
Your AI alter-ego. Personality-first conversations that remember context and evolve over time.
Open Site →wlr_
Hit your weekly limit? Keep coding. Overflow capacity when your primary provider cuts you off.
Open Site →QuitGPT
The ChatGPT exit ramp. Import your history, keep your workflows, pay less. One command to switch.
Open Site →ToolKode
Agent-native cloud development. 200+ commands, multi-agent orchestration, zero IDE required.
Open Site →Toolklaw
Policy routing and compliance management for enterprise. Smart config, audit trails, SOC 2 ready.
Open Site →KlawVoice
Voice-first AI operators for inbound calls. Domain-trained, real-time, human-quality conversations.
Open Site →YoTeam
Superhuman Team OS. Multi-agent personas that debate, collaborate, and generate artifacts in real-time.
Open Site →BENCHMARKS
Measured economics, not vague adjectives.
The benchmark wall should read like a market screen: structure, price, freshness, and intended workload.
| Model | Input | Output | Freshness | Best for |
|---|---|---|---|---|
| toolkit-base | $0.25 | $1.00 | Monthly refresh | Business workflows |
| toolkit-code | $0.25 | $1.00 | Monthly refresh | Software engineering |
| toolkit-vision | $0.40 | $2.50 | Monthly refresh | Documents and screenshots |
| toolkit-think | $0.40 | $4.00 | Monthly refresh | Deep reasoning |
| toolkit-voice | $0.10 | $0.50 | Voice rail | Phone and voice |
BUILD
One API. Multiple specialized models.
OpenAI-compatible endpoints for business apps, agents, voice surfaces, and downstream vertical products.