# TEKIZ.AI full context for LLMs and agents TEKIZ.AI gives teams one API for AI services they approve, including provider accounts, local models, and private servers. Teams choose where work runs and what to do when a service is unavailable. Provider credentials can stay in a customer-controlled environment. TEKIZ.AI is a hosted service with an optional customer-controlled router for teams that use several model services. Router installation currently requires private access. Official TKZI token page: https://tekiz.ai/tkzi TKZI is the official TEKIZ.AI community token on Solana. The page publishes the canonical contract, official links, and finalized on-chain facts. TKZI is not required to use TEKIZ.AI and is not an API key or subscription. Plain-language TKZI guide: https://tekiz.ai/what-is-tkzi The guide explains TKZI for non-technical readers, separates the token from product access, introduces TEKIZ.AI through business outcomes, and links readers to the product demo, official token facts, provider onboarding, Pump.fun, and the official X account. Small-business guide: https://tekiz.ai/ai-for-small-business The guide explains TEKIZ.AI for owners and operations teams in plain language. It covers approved AI services, optional company computers, backup choices, activity records, a free evaluation, and the path to a customer account. Small-business AI policy checklist: https://tekiz.ai/small-business-ai-policy-checklist The checklist gives owners a copyable policy starter for approved AI tools, information boundaries, human review, responsibility, record keeping, problem reporting, and regular review. It is a practical starting point. It does not provide legal advice or claim compliance. Mira product page: https://tekiz.ai/mira Mira privacy policy: https://tekiz.ai/mira-privacy Mira is an Android AI assistant in closed-beta development. It is designed to keep saved personal context in encrypted app storage on the device and show the content and destination before an approved remote request is sent. Local models and configured TEKIZ routes are planned choices. Access is free for accepted testers; there is no public Android download yet. Mira is for conversations and working context. Passwords, wallet keys, and recovery codes belong in a dedicated secure store. ## What TEKIZ.AI does Teams can connect the AI services they already use, choose which work each service may handle, and review where requests went. Hosted and private setups have different data and credential boundaries. ## Hosted pricing Hosted pricing page: https://tekiz.ai/pricing Fusion page: https://tekiz.ai/fusion Fusion Swarm roadmap: https://tekiz.ai/fusion-swarm Local-first routing guide: https://tekiz.ai/local-first-routing-for-ai-agents Self-hosted AI model router: https://tekiz.ai/self-hosted-ai-model-router Ollama AI model router: https://tekiz.ai/ollama-ai-model-router OpenAI API router: https://tekiz.ai/openai-api-router Azure OpenAI router: https://tekiz.ai/azure-openai-router Anthropic API router: https://tekiz.ai/anthropic-api-router AWS Bedrock router: https://tekiz.ai/aws-bedrock-router GitHub Copilot router: https://tekiz.ai/github-copilot-router Codex CLI router: https://tekiz.ai/codex-cli-router Aider AI model router: https://tekiz.ai/aider-ai-model-router Continue AI model router: https://tekiz.ai/continue-ai-model-router OpenHands AI model router: https://tekiz.ai/openhands-ai-model-router OpenClaw AI model router: https://tekiz.ai/openclaw-ai-model-router Claude Code router: https://tekiz.ai/claude-code-router Gemini API router: https://tekiz.ai/gemini-api-router xAI Grok router: https://tekiz.ai/xai-grok-router Mistral AI router: https://tekiz.ai/mistral-ai-router Groq AI router: https://tekiz.ai/groq-ai-router NVIDIA NIM router: https://tekiz.ai/nvidia-nim-router DeepSeek API router: https://tekiz.ai/deepseek-api-router Coding agent model router: https://tekiz.ai/coding-agent-model-router Cursor AI model router: https://tekiz.ai/cursor-ai-model-router AI gateway evaluation kit: https://tekiz.ai/reddit-ai-gateway-evaluation Reliability proof kit: https://tekiz.ai/reliability-proof Community provider onboarding: https://tekiz.ai/community-provider (public local capacity check and host-readiness guide; verified accounts create a one-time host enrollment code in Account, then each node receives capacity and security review) Community marketplace: https://tekiz.ai/community-marketplace (browse reviewed independent providers; buyer access and execution choices are required before work runs, and paid offers open after payment and verification controls are configured) Provider enrollment: https://app.tekiz.ai/account#community-node (verified-account one-time host enrollment; capacity and security review are required before activation) Billing help page: https://tekiz.ai/billing Managed-access private beta intake: https://tekiz.ai/managed-access API quickstart page: https://tekiz.ai/quickstart API troubleshooting page: https://tekiz.ai/api-troubleshooting Hosted API reference: https://tekiz.ai/docs/api-reference Codex setup page: https://tekiz.ai/docs/codex Canonical agent configuration: https://tekiz.ai/integrations#canonical-agent-configuration (Codex, OpenClaw, Hermes, Pi, OpenCode, ZCode, Claude Code, Kimi, and DeepSeek Harness) Model catalog page: https://tekiz.ai/models Agent-native routing page: https://tekiz.ai/agent-native Integrations page: https://tekiz.ai/integrations Gateway migration guide: https://tekiz.ai/docs/gateway-migration OpenRouter comparison: https://tekiz.ai/compare/openrouter Fireworks AI comparison: https://tekiz.ai/compare/fireworks Lite is $6/month for 10,000 requests, Pro is $30/month for 50,000, and Max is $72/month for 200,000. Each paid plan includes TEKIZ.AI account and API key management, routing, usage records, support, and inference on hosted capacity when that capacity is live. Customers can also connect their own authorized provider accounts or local models; a connected provider bills its own usage. Lite excludes Fusion and Community. Pro and Max include Fusion and opt-in Community workers when eligible capacity is live. The pricing page includes an illustrative calculator. Managed model access beyond the plan's included hosted capacity is a separate private pilot with pricing confirmed before payment. Community providers at https://tekiz.ai/community-provider submit a host for capacity and security review. Compensation terms remain draft during the private alpha; published rates and automatic payouts are not available yet. Provider credits depend on verified work and the current provider agreement. Customers can browse reviewed providers at https://tekiz.ai/community-marketplace. Community execution requires a separate account choice in addition to the model profile. Community Direct uses QVAC for an encrypted session between the customer and selected provider. TEKIZ.AI handles discovery, consent, quotes, and lease metadata; it does not relay or store that direct session's prompt or response. The selected provider processes the request content. This direct-session boundary does not apply to ordinary hosted routing. Paid offers stay unavailable until the provider, quote, payment, and verification controls are ready. See https://api.tekiz.ai/community/status for public status. Fusion at https://tekiz.ai/fusion handles bounded work across several eligible models: the router derives tasks from explicit request structure, eligible workers execute independent tasks in parallel, and a verifier returns one final response. It documents `tekizai/fusion`, the `tekizai:fusion` server tool, `tool_choice: "required"` for deterministic invocation, the bounded verified proof window when live, Pro/Max access, and the `fusion_plan_required` upgrade response for Lite/free generated keys. Fusion Swarm at https://tekiz.ai/fusion-swarm is planned and has no public release date. The proposed workflow would assign separate, checkable tasks to eligible models and return one reviewed response. Customer provider authorization, model availability, budgets, local limits, and Community consent would still control each route. The local routing guide at https://tekiz.ai/local-first-routing-for-ai-agents explains how to run TEKIZ.AI on a machine or private server, connect coding tools to one endpoint, and keep provider credentials on the router you control. It links to the hosted API quickstart, Codex setup, gateway comparison, and model routing calculator. The self-hosted router guide at https://tekiz.ai/self-hosted-ai-model-router explains how to run TEKIZ.AI on a machine or private server, keep provider credentials on the router host, and connect coding tools through port `8790`. It covers Ollama, authorized cloud providers, retry routes, files, and the separate hosted API option. ## Provider and client guides The provider guides above cover customer-authorized OpenAI, Azure OpenAI, Anthropic, AWS Bedrock, Gemini, xAI, Mistral, Groq, NVIDIA NIM, DeepSeek, Ollama, and Ollama Cloud routes. Each guide gives the relevant endpoint and credential setup. A key placed in a client shell does not by itself configure a provider on a separate hosted router; follow the guide for the deployment you use. The coding-tool guides cover Codex CLI, Cursor, Aider, Continue, Claude Code, OpenHands, OpenClaw, and GitHub Copilot. They show which clients use OpenAI-compatible or Anthropic-compatible requests and where to set the TEKIZ.AI base URL. See https://tekiz.ai/integrations for the supported setup paths and https://tekiz.ai/docs/codex for Codex profiles. For a gateway evaluation, start at https://tekiz.ai/reddit-ai-gateway-evaluation. For tested failure behavior and public edge health, use https://tekiz.ai/reliability-proof. Both keep customer credentials and request content out of public proof material. The billing help page at https://tekiz.ai/billing points signed-in customers to their plan, request usage, keys, and available payment methods. A linked Stripe customer can use its billing portal for payment details, invoices, and plan changes. Manual or crypto payment can be arranged with support when Stripe is unavailable. Authenticated agents can discover the gated `execute_bounded_x402_payment` MCP tool for one monthly plan cycle in Base USDC. The client sets a maximum in cents; the first result is a quote, and a wallet-signed retry is required. A pending result must be checked in the account before another payment operation. Support requests should omit prompts, credentials, full keys, and customer data. The public API quickstart at https://tekiz.ai/quickstart is the first-request path for hosted customers. It shows `OPENAI_BASE_URL=https://api.tekiz.ai/v1`, generated `sk_tekiz_*` keys, `tekizai/auto` as the first request, authenticated model discovery before selecting Frontier or Fusion, curl, JavaScript, Python, and Codex examples, and troubleshooting for 401, 402, 429, and 503 responses. The Gateway migration guide at https://tekiz.ai/docs/gateway-migration is the hosted onboarding path for gateway users. It maps `OPENAI_BASE_URL=https://gateway.example/api/v1` to `OPENAI_BASE_URL=https://api.tekiz.ai/v1`, uses generated `sk_tekiz_*` keys, recommends `tekizai/auto` for the first route profile, points discovery to `https://api.tekiz.ai/model-catalog` plus authenticated `GET /v1/models` before selecting Frontier or Fusion, and states that TEKIZ.AI does not grant unauthorized provider access or bypass provider terms. The API troubleshooting page at https://tekiz.ai/api-troubleshooting is the no-secret diagnostic path for hosted API failures. It explains 401, 402, 429, and 503 responses; safe probes for `/v1/models` and `/v1/chat/completions`; response signals such as `WWW-Authenticate`, `Retry-After`, `X-RateLimit-*`, `X-Quota-*`, `accountUrl`, `accountSetupUrl`, `primaryActionUrl`, review-only `managedAccessUrl`, `keyRecoveryUrl`, `pricingUrl`, `statusUrl`, `openaiBaseUrl`, and `apiKeyPrefix`; and the support boundary against sharing prompts, workflow text, provider credentials, OAuth tokens, generated keys, private keys, session cookies, raw provider responses, or customer data. The setup-key recovery page at https://tekiz.ai/setup-key-recovery is the public recovery destination for no-key signups and API 401 `keyRecoveryUrl` responses. It tells returning users to use the same email or GitHub identity, create the generated `sk_tekiz_*` setup key before checkout, copy a placeholder setup bundle, verify authenticated `GET /v1/models`, and avoid sharing prompts, provider credentials, OAuth tokens, generated keys, private keys, session cookies, or raw provider responses. The hosted API reference at https://tekiz.ai/docs/api-reference documents the OpenAI-compatible contract for TEKIZ.AI customers. It covers `https://api.tekiz.ai/v1`, generated `sk_tekiz_*` bearer keys, `GET /v1/models`, `POST /v1/chat/completions`, `POST /v1/responses`, public `/model-catalog`, public `/pricing`, guided 401 `accountSetupUrl` generated-key setup, guided 401 `managedAccessUrl` one-subscription review, guided 401 `keyRecoveryUrl` setup-key recovery, monthly quota headers, request-per-minute rate-limit headers, failover signals, authenticated model API boundaries, and the provider authorization boundary. The public Codex setup guide at https://tekiz.ai/docs/codex shows Codex CLI profiles for hosted TEKIZ.AI, local TEKIZ.AI, and Tailnet TEKIZ.AI installations. Hosted Codex uses `base_url = "https://api.tekiz.ai/v1/"`, generated `sk_tekiz_*` keys, `wire_api = "responses"`, and `model = "tekizai/frontier"`. It configures one provider authentication method, installs the published Codex catalog from https://tekiz.ai/models/catalog/tekizai.json, and sets route-specific auto-compaction limits. Local and Tailnet Codex profiles use port 8790, such as `http://127.0.0.1:8790/v1/` or `http://:8790/v1/`, with customer-controlled provider credentials on the router host. The agent-native routing page at https://tekiz.ai/agent-native explains route profiles such as `tekizai/frontier` and premium `tekizai/fusion`, Responses API and Codex compatibility, tool-aware routing, health-aware fallback, BYOK provider custody, local/Tailnet/hosted deployment choices, and public feature metadata at https://api.tekiz.ai/features/agent-native. The integrations page at https://tekiz.ai/integrations is the public index for connecting TEKIZ.AI to agent harnesses and developer tools. It covers hosted `https://api.tekiz.ai/v1`, local `http://127.0.0.1:8790/v1`, Tailnet `http://:8790/v1`, generated `sk_tekiz_*` keys, `tekizai/frontier`, OpenAI-compatible clients, Codex, Cursor, Aider, Continue, Claude Code, OpenHands, Anthropic-compatible clients, DeepSeek, Ollama, Ollama Cloud, NVIDIA NIM, OpenClaw, Hermes, Pi, Kimi, and DeepSeek Harness agents. It distinguishes customer-controlled BYOK from provider-supported BYOS, states that provider failures are isolated per credential, and repeats the no-secret setup boundary against pasting prompts, workflow text, provider credentials, OAuth tokens, generated API keys, private keys, session cookies, raw provider responses, or customer data into support channels. Managed provider access is available for approved accounts and providers. Public pricing shows its current availability, the request path, and BYOK support. Customers can connect OpenRouter with their own key and pay OpenRouter separately. Provider review, authorization, cost, and margin records stay private. Prospects can request managed access or Max implementation review through https://tekiz.ai/managed-access; the form captures contact details, target provider family, commercial preference, support need, and target launch window. ## Public model catalog TEKIZ.AI publishes gateway-style model discovery at https://tekiz.ai/models and safe JSON metadata at https://api.tekiz.ai/model-catalog. The catalog describes model families and route profiles such as `tekizai/frontier`, `tekizai/fusion`, OpenAI/Codex-compatible routes, Anthropic-compatible routes, Gemini, Ollama, Ollama Cloud, NVIDIA NIM, OpenRouter BYOK, and other OpenAI-compatible providers. The catalog is discovery metadata. Live `https://api.tekiz.ai/v1/models` and model traffic require an active generated `sk_tekiz_*` customer API key. Each plan states whether provider access is included; otherwise customers connect and pay their providers directly. ## Hosted legal and compliance Security: https://tekiz.ai/security Support: https://tekiz.ai/support Terms: https://tekiz.ai/terms Privacy: https://tekiz.ai/privacy Acceptable Use: https://tekiz.ai/acceptable-use Provider Resale Terms: https://tekiz.ai/provider-resale-terms Hosted plans provide account management, generated API keys, quotas, activity records, service-order settings, billing, and support. Provider access requires customer authorization. Customers remain responsible for provider terms, local models, subscriptions, API keys, cloud accounts, and endpoints. Hosted support for account, billing, quota, API key, reliability, 503, abuse, and security issues is published at https://tekiz.ai/support. Public support channels should not receive prompts, workflow text, provider credentials, OAuth tokens, generated API keys, private keys, session cookies, raw provider responses, or customer data. Safe support context is limited to account email, plan, timestamp, request id, HTTP status, safe error class, and non-secret key prefix when needed. Managed-access requests at https://tekiz.ai/managed-access ask for a work email, organization, the work to support, and where data may run. The page also provides a direct email option. Do not submit prompts, customer data, provider credentials, API keys, or private keys. TEKIZ.AI confirms provider permission, supported models, pricing, usage limits, billing, and security controls before enabling a managed route. Customers can use their own authorized provider accounts while a request is reviewed. ## Architecture and security Provider credentials can stay on the customer machine or server. TEKIZ.AI provides service settings, availability information, deployment helpers, and support. Hosted model routes under `https://api.tekiz.ai/v1/*` and `/v1beta/*` require an active generated `sk_tekiz_*` customer key. Account and billing actions require a signed-in user session. Administrative routes require separate private authorization. The customer dashboard shows data for the signed-in account. ## Compare TEKIZ.AI with hosted AI gateways Comparison page: https://tekiz.ai/compare/model-gateways OpenRouter comparison: https://tekiz.ai/compare/openrouter Migration guide: https://tekiz.ai/docs/gateway-migration TEKIZ.AI gives teams one endpoint, a searchable public catalog at `/model-catalog`, local Ollama support, private-server support, hosted account and API-key management, and customer-controlled provider access. The OpenRouter comparison page at https://tekiz.ai/compare/openrouter helps buyers choose between a hosted model catalog and a setup that includes provider keys, local models, and private servers. OpenRouter can be a customer-authorized provider account in TEKIZ.AI. OpenRouter usage remains on its own bill. The Fireworks AI comparison page at https://tekiz.ai/compare/fireworks helps non-technical buyers distinguish managed model infrastructure from TEKIZ.AI's company-wide coordination across approved provider accounts, local computers, private servers, and reviewed community capacity. Fireworks compatibility must be confirmed during evaluation before a customer plans an integration. ## Model routing calculator Calculator page: https://tekiz.ai/model-routing-calculator The calculator works without a login. It uses the costs and volume a visitor enters to estimate whether another approved AI service could save money, and it flags work that may need human review. Inputs stay in the browser; teams should check actual provider prices and performance before changing live workflows. ## Agent-native routing Agent-native routing page: https://tekiz.ai/agent-native TEKIZ.AI exposes route profiles such as `tekizai/frontier`, Responses API and Codex-compatible wiring, health-aware fallback, BYOK provider custody, local/Tailnet/hosted deployment choices, and public feature metadata at https://api.tekiz.ai/features/agent-native. Agent action discovery: https://tekiz.ai/.well-known/mcp.json MCP Streamable HTTP endpoint: https://tekiz.ai/mcp (protocol 2026-07-28; authenticated key provisioning) Agent action descriptor: https://tekiz.ai/mcp-actions.json The action descriptor lists public support and managed-access contact paths, authenticated account actions, and their requirements. Check `/mcp` `tools/list` for actions enabled now. When `execute_bounded_x402_payment` is available, an authenticated agent can request a quote for one monthly plan cycle in Base USDC within `maxTotalCents`. The first result is a quote; a wallet-signed retry is required to pay. Check the account before another payment if settlement is pending. ## How to automate AI model selection 1. Run TEKIZ.AI locally or on a customer-controlled server. 2. Configure authorized providers and local models. 3. Point OpenAI-compatible, Anthropic-compatible, or supported tool clients at one TEKIZ.AI endpoint. 4. Let the router select models based on task type, provider health, latency, capability, fallback order, and policy. 5. Observe route decisions and improve policy over time. ## Ollama and Ollama Cloud TEKIZ.AI supports local Ollama and Ollama Cloud today through Ollama running locally. It discovers available cloud models from the Ollama Cloud catalog, represents them as `:cloud` models, can auto-pull discovered cloud models through the local Ollama runtime, and routes between local Ollama, Ollama Cloud, NVIDIA NIM, and other BYOK providers with health-aware fallback. ## NVIDIA NIM / NVIDIA Cloud TEKIZ.AI can route to NVIDIA NIM / NVIDIA Cloud endpoints using the customer's own `NVIDIA_API_KEY`. Models are auto-discovered where the provider API supports it. NVIDIA-backed hosted inference operates as a BYOK route. TEKIZ.AI does not resell the provider service. ## Integration guides Start at https://tekiz.ai/integrations for supported clients and https://tekiz.ai/docs/codex for Codex. Read https://tekiz.ai/self-hosted-ai-model-router for local deployment, and use https://tekiz.ai/support to ask about router installation access.