Pay as you go
From $5 top-upprepaid credits, pay per tokenPublished per-token rates on every model, one balance per workspace and no seat fees. Credits last a year. No contractual SLA; priority email support. Official source 1 ↗
Website-checked tool profile
RunInfra
Hosted open-model API for coding agents, with cheap cached input and monthly coding plans from $10.
Sources checked
- Pricing
- Two ways to pay from the same workspace keys. Pay as you go uses prepaid credits (one credit is one US dollar, minimum top-up $5, credits last a year) at published per-token rates, for example $0.15 input, $0.01 cached input and $0.60 output per million tokens on DeepSeek V4.1 Flash 1017. Coding plans cost $10 (Starter), $29 (Pro) or $99 (Team) a month and include usage that resets every 5 hours and every week. New accounts get $1 of free credit once. Enterprise pricing is custom. Official source 1 ↗ · Official source 2 ↗ · Official source 3 ↗
- Available as
- An HTTP API at api.runinfra.ai with OpenAI-compatible chat completions and an Anthropic-compatible Messages endpoint, so existing SDKs and tools work by changing the base URL. · A web dashboard for keys, billing and plans, and the runinfra command-line tool (Node.js 22 or newer) that connects coding agents. Official source 1 ↗ · Official source 2 ↗ Official source 1 ↗ · Official source 2 ↗
- Primary group
- Coding & Development
Also consider
See all 6 similar tools ↓- ColdteaAn agentic development environment spanning code, release testing, and monitoring.Free plan
- ExaWeb search API for AI agents, with page contents, deep-research agents, monitors and an MCP server.Free plan
- InflightFeedback layer for prototypes that turns pinned comments and video into AI next steps your coding agent pulls over MCP.Free trial
What it does
RunInfra hosts open-weight language models, such as DeepSeek V4.1 Flash, GLM 5.3, Qwen3.8 and Ornith 1.5, behind one API key that works with the OpenAI and Anthropic SDKs. It is tuned for coding agents that resend the same long context on every turn: session-aware routing sends repeat calls to the server that already holds the cached prefix, and cached input is billed at a much lower rate. Developers pay per token from prepaid credits, or pick a monthly coding plan for Claude Code, Codex, OpenCode and other agents. RunInfra is run by RightNow AI, Inc., a Delaware company backed by Y Combinator.
Official source 1 ↗ · Official source 2 ↗Features
- Open models through OpenAI and Anthropic APIs. The pricing page lists eight hosted open models, including Ornith 1.5 122B, DeepSeek V4.1 Flash, GLM 5.3 Flash, Nemotron 3.5 Lightning 30B and Qwen3.8 27B, each with its own input, cached-input and output price. Models such as Ornith 1.5 122B support tool calling, JSON mode, streaming and image input. Official source 1 ↗ · Official source 2 ↗
- Session-aware prompt caching. Calls from the same session or conversation are routed back to the replica holding their cached prefix, automatically or via a session ID or prompt_cache_key. Cache hits show up in usage and are billed at the cached input rate. Hits are best effort. Official source 1 ↗ · Official source 2 ↗
- Published prices, speed and cache-hit rates per model. Each model page lists input, cached input and output prices per million tokens, measured output speed, time to first token, the last 24 hours' cache hit rate, context window and request limits. Official source 1 ↗
- Coding plans for coding agents. Monthly plans give a workspace included usage for Claude Code, Codex, OpenCode and 20+ other agents, metered at standard per-token prices against a 5-hour limit and a weekly limit. At a limit you choose between paying on with credits (with an optional cap) or the free, slower Standby mode. Official source 1 ↗ · Official source 2 ↗
- One-prompt setup inside your coding agent. You paste one prompt into your coding agent, approve the sign-in in your browser and confirm the setup plan; the agent then configures RunInfra through the CLI. Supported agents include Claude Code, Codex, OpenCode, Aider, Cline, Goose, Zed and others. Cursor and similar apps connect with an API key and base URL. Official source 1 ↗
- Zero data retention and SOC 2 Type II. Hosted model requests are zero data retention by default and are never used for training. RunInfra states it is SOC 2 Type II attested, with the report available under NDA. Official source 1 ↗ · Official source 2 ↗
- Spending controls and auto recharge. Owners can set a monthly spending limit per API key, mark keys as credits-only so they skip the plan, and turn on auto recharge with a threshold and a monthly ceiling. Official source 1 ↗ · Official source 2 ↗
Pricing & free limits
Coding plan Starter
$10per monthAbout $12 of usage a month at RunInfra's rates: $1.11 per 5-hour window and $2.77 per week, plus up to $0.20 a day of Standby. Official source 1 ↗
Coding plan Pro
$29per monthAbout $35 of usage a month: $3.23 per 5-hour window and $8.08 per week, plus up to $0.58 a day of Standby. Official source 1 ↗
Coding plan Team
$99per monthAbout $119 of usage a month: $10.98 per 5-hour window and $27.46 per week, plus up to $1.98 a day of Standby. Limits are shared by the workspace, not per member. Official source 1 ↗
Enterprise
CustomcontractDedicated infrastructure, audit logs and RBAC, custom volume and terms, SLAs up to 99.99%, SOC 2 Type II report, and a dedicated success manager with private Slack. Official source 1 ↗
Billing details
Coding plans renew monthly until cancelled. Cancellation takes effect at the end of the billing period; upgrades apply immediately with a prorated charge, downgrades at period end. Plan usage does not roll over. Official source 1 ↗ · Official source 2 ↗
Purchases are non-refundable. The only exception is a coding plan that never starts serving, which is refunded in full. Deleting your account ends the plan at once with no refund for the current month. Official source 1 ↗ · Official source 2 ↗
Top-up credit expires one year after purchase and is spent soonest-expiring first. Images, audio and other non-chat work always bill from credits, not from a coding plan. Official source 1 ↗ · Official source 2 ↗
Free access & limits
Every new account gets $1 of free credit once, which never expires, to test the Model APIs. Official source 1 ↗ · Official source 2 ↗
On a coding plan, the Standby mode keeps chat requests running free after a limit, but slower and best effort: it refills at midnight UTC and can be reduced or paused at any time. Official source 1 ↗
Best for & limitations
Running Claude Code, Codex or OpenCode on cheaper open models under a fixed monthly coding plan, set up by pasting one prompt into the agent. Official source 1 ↗ · Official source 2 ↗
Cutting input costs for agent workloads that resend the whole conversation on every tool call, since the repeated part is served from cache at the cached rate. Official source 1 ↗
Pointing an existing app built on the OpenAI or Anthropic SDK at open models by switching the base URL and key, paying per token from a prepaid balance. Official source 1 ↗ · Official source 2 ↗
- Cache hits are not guaranteed. The low cached prices depend on cache hits, which RunInfra describes as best effort and lost on eviction or server restart. Official source 1 ↗
- Small included usage on lower plans. Starter covers only $1.11 of usage per 5-hour window and $2.77 per week; beyond that you pay with credits or drop to the slower Standby mode. Official source 1 ↗
- Non-refundable purchases. Credit purchases and plan payments are not refunded, except for a coding plan that never starts serving. Official source 1 ↗
- No single sign-on or self-serve SLA. SAML or OIDC single sign-on is not offered, and pay-as-you-go accounts have no contractual SLA; SLAs come only with Enterprise. Official source 1 ↗ · Official source 2 ↗
- Not for health data. The service is not currently offered for processing protected health information. Official source 1 ↗
Similar tools
Tools like RunInfra
Published tools that share its categories.
ColdteaAn agentic development environment spanning code, release testing, and monitoring.
Visit ExaWeb search API for AI agents, with page contents, deep-research agents, monitors and an MCP server.
Visit InflightFeedback layer for prototypes that turns pinned comments and video into AI next steps your coding agent pulls over MCP.
Visit 
Gemini 4 ArgonGoogle’s flagship Gemini 4 reasoning model for hard coding, agent and research tasks, with a 1M-token window.
Visit 
Claude Opus 5.5Anthropic’s Opus-tier Claude model for long-running agentic coding and knowledge work, at $4/$20 per million tokens.
Visit 
OpenCodeOpen-source AI coding agent for the terminal, desktop and IDE that works with free models, 75+ providers or your existing subscriptions.
Visit