SentioGate

Zero-markup BYOK AI Router and FinOps Control Plane. Audit past LLM token waste, track live spend and prompt-cache savings, pre-flight check agent tool safety, and route coding tasks to cost-optimal models.

View on sentiogate.app

Documentation

SentioGate

SentioGate (https://sentiogate.app) is an AI LLM router and unified LLM API gateway: one OpenAI-compatible API for every model, on your own provider keys (BYOK) at 0% markup, running on a global edge network. It is distinct from Sentio.xyz (Web3 data indexer) and Sensogate (dental care). Change one base URL to reach OpenAI, Groq, DeepSeek, Mistral, xAI, Claude and Gemini through OpenRouter, or your own HTTPS vLLM/Ollama endpoint. SentioGate picks a capable model per request automatically, automatically falls back to another model if one fails, trims long conversations so you spend fewer tokens, and stops runaway agent loops. There is a free tier to start.

Public Pages & Documentation

  • Home: Interactive routing playground, drop-in code examples, and live telemetry overview
  • Free Chatbot API: Free Tiers Compared: a sourced comparison of free chatbot API tiers (Gemini, Groq, Mistral, OpenRouter, Hugging Face), plus an OpenAI-compatible key on the free community pool (send model "auto"; 600 requests per 5 hours, no credit card) with cURL, Python, website and Ollama examples
  • Architecture & Alternatives Comparison: SentioGate (0% markup BYOK edge gateway) vs. OpenRouter, LiteLLM, and Portkey
  • Guides: Dated, sourced how-to guides for multi-provider LLM apps
  • OpenRouter free model limits and 429s: OpenRouter's :free limits, what counts toward the cap, how to check remaining requests, why extra keys do not help, and cross-provider fallback with working code
  • Route OpenCode through an LLM router: Tested opencode.json for an OpenAI-compatible router, and Claude Code setup through SentioGate's Anthropic Messages endpoint (/v1/messages)
  • Best OpenRouter alternatives (2026): Requesty, SentioGate, LiteLLM, Portkey, Helicone and direct APIs like Groq compared with sourced fees and free tiers, and when to stay on OpenRouter
  • OpenRouter vs direct API: What OpenRouter's fee on credits costs vs paying providers directly, what it buys, BYOK, and when to go direct
  • Use SentioGate with LangChain: ChatOpenAI setup in Python and JavaScript with SentioGate's base URL, streaming, tools and session ids
  • Use SentioGate with Langfuse: Opt-in trace export to your own Langfuse project (metadata only by default) or the Langfuse OpenAI SDK wrapper
  • SentioGate vs OpenRouter: Fees, BYOK, free tiers, routing and when to pick each, with every OpenRouter fact linked to its official page (as of Oct 30, 2026)
  • OpenRouter vs Groq: Multi-provider router vs direct LPU inference API: models, billing, rate limits and 429 handling (as of Oct 30, 2026)
  • OpenRouter vs Requesty: Two hosted LLM routers compared on fees, BYOK, free tier, routing policies and governance (as of Oct 30, 2026)
  • OpenRouter vs LiteLLM: Hosted model marketplace vs self-hosted gateway: fees, BYOK, routing, governance and where it runs (as of Oct 30, 2026)
  • OpenRouter vs Portkey: Model marketplace vs AI gateway: credit fees vs log-based plans, BYOK, fallbacks, governance and self-hosting (as of Oct 30, 2026)
  • OpenRouter vs Vercel AI Gateway: Two hosted gateways compared on fees, BYOK rules, free tiers, fallbacks and spend controls (as of Oct 30, 2026)
  • OpenRouter vs Together AI: Multi-provider router vs direct inference provider: billing, free tier, API compatibility and regions (as of Oct 30, 2026)
  • OpenRouter vs DeepInfra: Multi-provider router vs per-token inference provider: pricing, concurrency limits, OpenAI compatibility and fallback (as of Oct 30, 2026)
  • OpenRouter vs Fireworks AI: Router vs inference platform: prepaid credits, serverless and GPU pricing, rate limits and spend limits (as of Oct 30, 2026)
  • OpenRouter vs Hugging Face Inference Providers: Two LLM routers compared on markup, credits, BYOK, provider selection and failover (as of Oct 30, 2026)
  • OpenRouter vs Amazon Bedrock: Router vs AWS model service: pricing tiers, OpenAI-compatible API, IAM access, quotas and routing (as of Oct 30, 2026)
  • Portkey alternative: Simple hosted BYOK router vs Portkey's guardrails, MCP and enterprise gateway, and when to pick Portkey
  • OpenAI 429 errors: retry and fallback: rate_limit_exceeded vs insufficient_quota, retry with Retry-After and jitter, and cross-provider fallback with working code
  • Helicone alternative: When you need a lightweight hosted BYOK router vs Helicone's full LLM observability platform
  • BYOK LLM gateway explained: How Bring-Your-Own-Key routing works at 0% token markup vs credit resellers and self-hosted proxies
  • OpenAI to Claude & multi-provider failover: How to fail over across providers on HTTP 429 and 5xx errors
  • Gemini 429 RESOURCE_EXHAUSTED errors: Per-project RPM/TPM/RPD limits, Retry-After backoff, and cross-provider fallback
  • Groq 429 rate limit errors: Groq's retry-after and x-ratelimit headers and instant multi-provider failover
  • How to reduce LLM API costs: Cost-first model routing, context compression, semantic caching and per-key budgets
  • LLM semantic caching: Provider prefix caching vs gateway semantic caching, workspace isolation and opt-out controls
  • LLM router hub: What an LLM router does, free and BYOK options, hosted vs self-hosted, and links to every comparison and guide
  • What is an LLM gateway?: Plain-English definition of an LLM gateway and how it differs from an LLM router or proxy
  • Free LLM API list: Which LLM APIs are free, trial-only or retired, each checked on the provider's official page (Oct 2026)
  • Best free chatbot API: How to pick a free chatbot API and build a chat loop that handles rate limits
  • LiteLLM vs Bifrost vs Helicone vs SentioGate: Sourced gateway comparison dated 2026-10-20, including SentioGate's limitations
  • FAQ: Grouped answers on the free tier, pricing and BYOK, routing and failover, and integrations, with links to the page that covers each question in depth
  • Live LLM Benchmarks: Public rolling-average TTFT, p50/p95 latency, HTTP 429 rate-limit tracker, and outage simulator
  • Pricing & Free Beta: Zero-markup BYOK plans and the 90-Day Free Beta
  • API Reference: Interactive OpenAPI documentation (machine-readable spec: https://api.sentiogate.dev/openapi.json)
  • Live Benchmarks JSON API: Machine-readable provider latency and error-rate telemetry
  • Model Context Protocol (MCP) Server: Streamable HTTP MCP server and stdio package (@sentiogate/mcp, Registry ID: io.github.SentioGate/sentiogate-mcp, GitHub: https://github.com/SentioGate/sentiogate-mcp) exposing FinOps waste auditing, tool safety guards, and cost-optimal model recommendations to Claude Code, Cursor, Cline, OpenCode, and the Nous Hermes Agent harness
  • Senti Bot & Autonomous Coding Dashboard (sentiogate.cloud): Built on the Nous Hermes Agent harness (NousResearch/hermes-agent) with SentioGate Edge Gateway & MCP Control Plane (@sentiogate/mcp)

Core Capabilities

  • Zero-Markup BYOK Key Vault: Provider API keys are encrypted at rest; supports up to 5 keys per provider in priority failover order, with optional per-key spend and balance limits that pause a key.
  • Automatic Model Choice & Fallback: model: "auto" picks a capable model for the task at a good price, skips models that are failing or unusually slow, and automatically falls back to another model if one fails. Pin any provider/model to choose it yourself.
  • Leaner Long Conversations & Response Reuse: Trims long conversations so you spend fewer tokens, and reuses answers to repeated requests (opt out per request).
  • Tool-Output Trimming for Agents: Shortens large tool results from earlier agent steps; the newest results and user messages are never changed. On by default; opt out per request with header X-SentioGate-Tool-Trim: off.
  • Runaway Loop Warning: Flags an agent that looks stuck repeating the same steps in the X-SentioGate-Loop response header and, by default, answers with a 400 loop_detected error instead of a model call; send X-SentioGate-Loop-Stop: off (or turn off "Stop runaway loops" in the dashboard) to only flag it.
  • Dangerous Tool-Call Flag: Flags tool calls in the model's answer that look dangerous (deleting the home directory, piping a download into a shell, sending credential files or secrets out, force-pushing to main) in the X-SentioGate-Guard response header (an SSE comment on streams) and router.guard. Flag only, the answer is never changed; send X-SentioGate-Tool-Guard: off (or turn off "Flag dangerous tool calls" in the dashboard) to skip it.
  • Per-Request Controls: Set route_preference (cost, quality or latency) and task_class in the request body, keep a conversation on the same route with X-Session-ID, and switch features per request with X-SentioGate-Cache, X-SentioGate-Tool-Trim, X-SentioGate-Loop-Stop and X-SentioGate-Tool-Guard. Trimming, loop stop and the tool-call flag also have account-wide switches under Routing policy in the dashboard.
  • Usage & Cost Analytics: Dashboard with requests, spend at your providers' prices, cache savings, latency, success rate, daily requests and recent routing decisions (also via GET /v1/analytics and GET /v1/decisions). Every response carries router metadata (chosen model, estimated cost, attempts). Optional Langfuse trace export, metadata only by default.
  • Admin & Key Management: Create, list and revoke gateway API keys; up to 5 provider keys per provider in fallback order; turn individual models on or off; daily spend and requests-per-minute caps per provider and per model; per-key spend and balance limits that alert or pause a key; your own custom OpenAI-compatible providers.
  • Background Queue (advanced): For background jobs only, a non-streaming request on the free community pool can send header X-SentioGate-Allow-Queue: on to be queued (202 Accepted with a poll link) instead of getting a busy error.

Policies