Get your API key
Sign in, create an API key, and try your first 10 requests free. No card required.
- Account
- API key
- Access
- Try it
We will email you a sign-in link.
Pro Create key firstMore options
Choose a plan
Your selection is saved and applied after sign-in.
See the client setup
Use password sign-in instead
Account
Loading...
Create an API key, check that it works, then send your first request.
Setup progress
See what is left to do. A key check confirms access, and a test request confirms TEKIZ.AI is working.
Sign in to manage your keys and plan.
Create at least one active sk_tekiz_* key.
Claim 10 free requests or choose a plan to activate routing.
Check /v1/models before connecting an app.
Send a test request from this browser.
Connect a client
Your first request worked. Copy the Codex setup or choose another supported client.
Account details
Current usage and a support-safe account summary.
Usage
Loading current billing period...
Share with support
Copy a short account summary for support. It includes your plan, usage limits, and account status. It never includes prompts, credentials, API keys, cookies, model responses, or customer data.
API key
Your newly generated key appears here once.
Connect your app
OPENAI_BASE_URL=https://api.tekiz.ai/v1
ANTHROPIC_BASE_URL=https://api.tekiz.ai
Hosted models depend on your plan. To use your own provider keys, run TEKIZ.AI on a machine or server you control.
Run a Community node
Offer a model running in Ollama or an OpenAI-compatible server such as vLLM. Standard hosts can start automatic readiness verification during setup. Admission must succeed before a node can accept work.
Your nodes
Loading node status...
Accepted work
Loading accepted work...
Accepted jobs earn compute credits when an earning policy is enabled. Your credit balance and recent activity appear below.
Compute credits
Loading compute credits...
Credits bank accepted contributions for future inference. Redemption depends on enabled model rates, your saved execution and residency consent, and available capacity.
For hardware and Ollama subscription workers, use Linux with an active systemd user session, Python 3, Bash, curl, jq and tar. Hardware workers also need Ollama and an installed model that fits available memory. Download the worker on the computer you want to contribute. Use a new folder and keep it in place while its service is installed.
mkdir -m 700 tekizai-community-setup && cd tekizai-community-setup && curl -fSLo community-worker-18bf8d49cd22.tar.gz https://tekiz.ai/downloads/community-worker-18bf8d49cd22.tar.gz && printf '%s %s\n' '8e22d9dde6c114402ee241478c384b32e6132d7fac4ad7e582ea10897eefae33' 'community-worker-18bf8d49cd22.tar.gz' | sha256sum -c - && tar -xzf community-worker-18bf8d49cd22.tar.gz && cd tekizai-community-worker
Run the commands below from the downloaded worker folder. Keep provider keys on your computer. Hardware setup and recovery instructions are included in docs/community-contributor-quickstart.md; subscription instructions are in docs/community-ollama-subscriptions.md.
Choose a host type
Check one model
This reads local model server and hardware details without loading a model, changing files, or contacting TEKIZ.AI.
A node runs one job and one loaded model at a time. We use measured RAM and VRAM to review its capacity. The default model limit is 34B parameters; we can raise it after a capacity review.
Request access
Create a temporary setup code, then run the enrollment command in the downloaded worker folder on Linux with an active systemd user session. It installs and starts the worker service and requests readiness verification. Work becomes available after admission succeeds.
Check your node service
The enrollment command starts the worker as a Linux user service. Keep it running for readiness checks and eligible work. The worker makes outbound requests only.
systemctl --user status tekizai-router-community-node.service
Community routing consent
Choose which Community modes and regions this account may use. If you leave all options unselected, Community routing stays off. Plan and node security checks still apply, and every change is recorded.
Community operators may see plaintext while executing non-TEE work. Member Ollama subscriptions also send requests to Ollama. To use them, allow hosted provider models, non-TEE workers and Global hosted processing. This permits provider-selected processing locations. Your plan and capacity restrictions still apply.