chutes-agent-toolkit

Chutes.ai hub skill — the entry point for decentralized open-source AI inference on Chutes. Use this skill when a user mentions Chutes, chutes.ai, decentralized AI, DeepSeek/Qwen/Kimi/GLM API access, TEE or confidential-compute inference, serverless GPU inference, or wants to register, manage API keys, discover models, or make OpenAI-compatible inference calls. This is the hub: it handles the session credential check, account lifecycle, API key lifecycle, basic inference, and routes to sibling skills (chutes-sign-in, chutes-deploy, chutes-mcp-portability, chutes-routing, chutes-usage-and-billing, chutes-platform-ops, chutes-agent-registration) for everything else.

Make Chutes.ai available to any agent via MCP or OpenAI-compatible configs. Use this skill when the user wants to plug Chutes into Cursor, Cline, Aider, Claude Desktop, Hermes, or another agent. The skill installs the Chutes MCP server (stdio), writes per-agent config snippets, and runs a health check. Triggers on: MCP chutes, chutes MCP server, cursor chutes, cline chutes, aider chutes, claude desktop chutes, hermes chutes provider, openai-compatible chutes, drop-in chutes config, portable chutes, chutes for my agent.

Chutes.ai model routing and pooling. Use this skill when the user wants to set up model failover, latency or throughput routing, TEE-only filters, inline model pools, or stable model aliases for Chutes. Also use when picking a pool for a specific intent (interactive-fast, cheap-background, private-reasoning, agent-coder, tee-chat). Triggers on: chutes routing, model pool, default:latency, default:throughput, TEE-only routing, failover, model alias, /model_aliases/, chutes routing recipe, build a pool, alias pack.

[BETA] Chutes.ai autonomous agent registration from Hermes: check registration status, prepare a Bittensor-backed agent account registration, and surface on-chain/coldkey prerequisites. Triggers on: chutes agent registration, register chutes agent, autonomous agent chutes, bittensor agent, coldkey, agent status.

Use Chutes.ai from Hermes for model discovery, routing, account/API-key guidance, and OpenAI-compatible inference. Prefer the live models endpoint for current availability and capabilities.

[BETA — permanent until verified] Deploy models on the Chutes.ai decentralized GPU network from Hermes. vLLM / diffusion / custom CDK deploy, image builds, teeify, rolling updates, stable aliases. Triggers on: deploy chute, chutes deploy, vllm chute, diffusion chute, build chute image, POST /chutes/, teeify, rolling update chutes, model alias deploy.

[BETA] Make Chutes.ai available to Hermes and any other MCP or OpenAI-compatible agent. Installs the Chutes MCP server (stdio), writes drop-in configs for Cursor / Cline / Aider / Hermes custom-provider / system-prompt agents, and runs a health check. Triggers on: MCP chutes, chutes MCP server, hermes chutes provider, cursor chutes, cline chutes, aider chutes, openai-compatible chutes, drop-in chutes config.

Chutes.ai platform operations for Hermes users: OAuth app fleet audits, bulk secret rotation dry-runs, alias CRUD/audits, token introspection and revocation guidance. Triggers on: chutes platform ops, oauth fleet, audit stale apps, rotate all secrets, alias crud, revoke token, introspect token, /idp/apps, /model_aliases/.

Chutes.ai model routing and pooling for Hermes users: build live model pools, latency/throughput routing strings, TEE-only filters, and stable aliases. Triggers on: chutes routing, model pool, default:latency, default:throughput, TEE-only routing, failover, model alias, /model_aliases/, build a pool, alias pack.

[BETA] Add Sign in with Chutes (OAuth 2.0 + OIDC + PKCE) to a Hermes user's application. Registers OAuth apps via /idp/apps, vendors chutesai/Sign-in-with-Chutes Next.js files, wires server-side token handling, and manages client-secret rotation. Triggers on: sign in with chutes, SIWC, OAuth app, /idp/apps, cid_, csc_, PKCE chutes, chutes login button, rotate oauth secret.

Chutes.ai TEE and attestation workflow for Hermes users: fetch evidence envelopes, parse TDX quotes, parse NVIDIA GPU attestation, and explain what TEE does and does not guarantee. Triggers on: chutes tee, confidential compute, TEE attestation, fetch evidence, verify quote, GPU attestation, tdx quote, /chutes evidence.

Read-only Chutes.ai usage, quota, discount, and billing diagnostics for Hermes users. Triggers on: chutes usage, chutes billing, chutes quota, chutes discount, chutes balance, chutes spend, subscription usage, payments summary, cost breakdown, monthly usage, quota guard.

Chutes.ai agent-native onboarding. Use this skill when an autonomous agent needs to self-register on Chutes without a human signup, bootstrap its own Chutes account, and set up its first API key programmatically via a Bittensor hotkey signature. Triggers on: chutes agent registration, agent signup, agent_registration, agent setup, programmatic chutes signup, hotkey agent register, autonomous chutes onboarding, bittensor agent account.

[BETA — permanent until verified] Deploy models on the Chutes.ai decentralized GPU network. Use this skill when the user wants to deploy a vLLM chute, diffusion chute, custom CDK chute, build a chute image, deploy a TEE/confidential chute (tee=True), inspect rolling updates, or create a stable model alias after deploy. Triggers on: deploy chute, chutes deploy, vllm chute, diffusion chute, build chute image, POST /chutes/, POST /chutes/vllm, POST /chutes/diffusion, POST /images/, teeify, tee=True, private chute, deployment fee, chutes deploy --accept-fee, revision pinning, rolling update chutes, chute share/unshare/make_public, model alias deploy.

Chutes.ai platform operator flows: OAuth app inventory + audits, client-secret rotation at fleet scale, token introspection and revocation, authorization management, model alias CRUD as an operator task. Use when the user asks to audit their OAuth apps, rotate secrets across many apps, introspect or revoke tokens, inspect authorizations, or manage model aliases as a team-level operator. Triggers on: chutes oauth app audit, list oauth apps, stale oauth apps, rotate all secrets, token introspect, token revoke, authorizations, /idp/apps, /idp/token, alias CRUD, fleet operator, chutes doctor.

Add Sign in with Chutes (OAuth 2.0 + OIDC + PKCE) to a user's application. Use this skill when the user wants to register a Chutes OAuth app, vendor the chutesai/Sign-in-with-Chutes Next.js package into their project, wire up sign-in routes, manage scopes, store client credentials in the OS keychain, or rotate a client secret. Triggers on: sign in with chutes, SIWC, OAuth app, /idp/apps, cid_, csc_, chutes oauth, PKCE chutes, chutes login button, useChutesSession, Next.js chutes auth, rotate oauth secret, chutes oauth scopes, /idp/scopes, chutes:invoke, token introspect, chutes OIDC, openid-configuration.

Verify Chutes.ai TEE (Trusted Execution Environment) attestation evidence — Intel TDX quotes and NVIDIA GPU confidential-compute reports (Hopper H200, Blackwell B200/B300/RTX Pro 6000). Use this skill when a user wants to actually verify that a chute marked confidential_compute:true is really running inside hardware isolation, fetch attestation evidence from the chute endpoint, parse a TDX quote byte-level, inspect the GPU attestation report, compare measurements against Chutes' published golden values, or audit the certificate chain. Triggers on: TEE verification, confidential compute, TDX quote, Intel TDX, NVIDIA Hopper attestation, NVIDIA Blackwell attestation, GPU confidential compute, /chutes/{id}/evidence, /servers/tee/measurements, golden measurements, quote verification, DCAP, attestation, chute attest, tee_type, mrsigner, mrtd, rtmr.

Chutes.ai usage, quotas, discounts, and billing diagnostics. Use when the user asks about Chutes spend, balance, top-ups, discounts, quota usage, subscription usage, invocation stats, payment history, or 'why is my bill high'. Triggers on: chutes usage, chutes billing, chutes quota, chutes discount, chutes balance, chutes spend, invocations stats, subscription usage, payments summary, spend audit, cost breakdown, monthly usage, four_hour burst, quota guard.