Decentralized inference.
Private by design.

taoCode connects your coding agent to independent Bittensor gateways. Choose TEE-capable models, keep tools local, and decide where your inference runs.

Explore private compute ↓Get started →
taocode

Already installed? Run in your project. Public installer coming soon.

TEE-only by defaultUnderstand verification ↗
CONFIDENTIAL COMPUTECONCEPT / 01
Confidential compute
Hardware isolation for AI workloads.
14
models report TEE supportBrowse the model catalog
↗
How taoCode handles your data
Explore taoCode Privacy & TEEWorkflowGatewaysModelsFAQ
⌘

Choose your gateway

Select the gateway that receives your prompts and code. Switch models from the same terminal session.

⬡

TEE-first model selection

TEE-only mode restricts TAO models to those marked confidential-compute capable. Local models remain available.

✓

Check attestation status

See whether a gateway reports TEE support and whether the client’s attestation checks have passed.

Privacy & security

How taoCode handles
your code and data.↗

Local tools, remote inference, and hardware attestation each handle a different part of the task.

01Your context

Know what leaves your machine.↗

File edits and shell commands run locally. Your prompts, relevant code, and tool results go to the selected model endpoint for inference.

Local tools · remote inference
See what leaves your machine ↗
02Confidential compute

Select models with TEE support.↗

A Trusted Execution Environment isolates a workload in hardware. TEE-only mode is on by default for TAO models, restricting selection to models marked as confidential-compute capable.

14 models currently report TEE support
Understand TEE protection ↗
03Check the evidence

Check the attestation evidence.↗

With supported Chutes wallet authentication, taoCode checks hardware attestation locally. The status shows whether those checks passed; a gateway’s TEE label alone does not count as verification.

Claimed / verified / failed
See the verification checks ↗
PRIVACY LAYERSCONCEPT VISUAL
⌘
Your workspaceCode · prompts · tool results
TEE boundary
Attestation status
ClaimedVerifiedFailed
Reported by the client
01Your context

Illustration of the inference and attestation flow.

TEE capability is gateway-reported. Actual protection depends on the endpoint, transport, and verification flow.

Chutes’ privacy & evidence guide ↗
Transport & verification

Follow the request and the evidence↗

Transport encryption protects data in transit. Hardware attestation checks evidence about the environment running a workload.

Secure transport↗
  1. Client
  2. Encrypted connection
  3. Gateway
TLS encrypts traffic between your client and the gateway. The gateway terminates that connection. Encryption through to an enclave requires a separate supported transport.
TEE attestation↗
  1. Fresh nonce
  2. Signed quote
  3. Local checks
The client sends a fresh nonce and checks the signed quote returned for an enclave instance. taoCode supports these checks through Chutes wallet authentication.
See what taoCode verifies →
How it works

From prompt to subnet and back↗

Select a step to follow the request. Hover over the diagram to pause it.

tao/chutes/deepseek-ai/DeepSeek-V3.2-TEE
01 Your prompt
Gateways

One agent, many gateways↗

Chutes SN64

OpenAI-compatible inference with a public model catalog and per-model confidential-compute metadata.

Explore gateway →
Targon SN4

Confidential compute infrastructure on Bittensor, built for private AI workloads.

Explore gateway →
Model picker

Which model should I use?↗

Find a model for your next task. Choose a starting point and copy the command.

DeepSeek-V3.2 Chutes · SN64 FP8 · 128K ctx

Tool calling reported by gateway

taocode run --model tao/chutes/deepseek-ai/DeepSeek-V3.2-TEE "add pagination to the users endpoint"
Documentation

Set up, connect, and run a task↗

Installation steps, model selection, CLI commands, and configuration examples.

Open the docs ↗
FAQ

Common questions↗

Where does my code go?
The agent runs locally, but prompts and relevant file content are sent to your selected gateway for inference. Data handling depends on that gateway and endpoint. Read the privacy guide →
Link to this answer ↗
What runs on my machine?
The agent session and its tools run on your machine. The selected gateway runs model inference and returns responses and tool requests.
Link to this answer ↗
Can I switch gateways?
Yes. Connect a second gateway and select one of its model references. Catalogs, capabilities, access rules, and billing vary by gateway. Choose a model →
Link to this answer ↗
Do I need a crypto wallet?
No. You can connect Chutes or Targon with an API key. Chutes also supports wallet authentication for attestation checks.
Link to this answer ↗
Which models can I use?
taoCode includes a curated catalog and discovers models from connected gateways. Run taocode models for your available choices, or browse the model catalog.
Link to this answer ↗
How is inference billed?
Inference is billed by the gateway you connect. Check its pricing, account balance, and usage limits before starting a session.
Link to this answer ↗
Get started

Run your first
coding task↗

Connect a gateway, select a model, and ask taoCode to make a change in your project.

taocode
Get started → Explore the network