Token Harbor
Token Harbor is a unified API gateway that gives developers one OpenAI-compatible endpoint for Anthropic, OpenAI, Gemini, Zhipu, Kimi and other leading models. Usage-based token billing sits alongside free monthly allowances and prepaid monthly Passes, and the service promises that the model ID you send is the model that runs, with data-retention controls you set yourself.
What is Token Harbor?
Token Harbor is a unified LLM API gateway that exposes Anthropic, OpenAI, Gemini, Zhipu, Kimi and other leading models through a single OpenAI-compatible API, operated by Token Harbor PTE. LTD. Rather than holding separate accounts, keys and balances with each frontier lab, a developer signs up once, receives one endpoint and one key, and changes only the model ID to move between providers. The homepage states the promise plainly: one API for the world's leading AI models, with Anthropic, OpenAI, Gemini, Zhipu and Kimi named as the first families available.
The product is organised around four of its own stated promises. Model integrity means the model ID you pass is exactly what runs, with no silent substitution of a cheaper model behind your back. Data privacy means no baseline retention policy is applied behind the scenes: retention is yours to set based on your needs. Full transparency means the routing of every call, and where it runs, is explained rather than buried in fine print. A fourth promise covers responsive, human support when something goes wrong.
Access follows a free-to-paid ladder. The Free tier carries a monthly allowance and a rotating lineup that currently includes DeepSeek V4.1 Flash, MiMo V2.6 Flash and TH-Rudder, with promotional models added over time, and no card is required to start. Above it sit three Passes, billed monthly at $1.99 for Agent Pass after a $0.99 first month, $9.99 for Office Pass and $99 for Frontier Pass. Each bundles included usage value plus the free allowance, a model boost on selected models, and a standing discount on pay-as-you-go calls past the included amount. Passes spend on the standard per-token routes, which are not kept for training and are not stored beyond a five-minute exact-match cache.
Beyond the raw endpoint, Token Harbor ships a chat interface called Rudder, a models catalogue, documentation, a connect flow that wires Claude Code, Cursor or Codex to the gateway in one command, a status page, a blog and a Tharbor area. The same account therefore covers both programmatic traffic and hands-on evaluation, and because billing is measured at published per-token prices, teams can estimate request volumes per model before choosing a pass.
Token Harbor Core Features
Single OpenAI-compatible endpoint
one key and base URL reach Anthropic, OpenAI, Gemini, Zhipu, Kimi and other leading model families.
Model integrity guarantee
the model ID you pass is exactly the model that runs, with no cheaper substitution behind your back.
User-set data retention
no baseline policy is applied behind the scenes, so you decide how prompts are treated.
Transparent routing
the site explains how every call is routed and where it runs instead of hiding the infrastructure.
Free monthly allowance
a recurring pool covering DeepSeek V4.1 Flash, MiMo V2.6 Flash and the TH-Rudder model, with promotional models added over time.
Passes with included usage
Agent, Office and Frontier bundle $10, $35 and $180 of usage value, boosted to $20, $70 and $360 on selected models.
Standing overage discounts
passes take 5%, 10% or 15% off pay-as-you-go calls billed from your balance past the included value.
Chat and agent connect
use the Rudder chat interface for hands-on work or point Claude Code, Cursor and Codex at the endpoint.
Who is Token Harbor for?
Token Harbor suits developers and small product teams who need several frontier models without opening several vendor accounts. The typical buyer already writes code against an OpenAI-compatible client, so changing a base URL and key is a far smaller job than negotiating terms with each lab. Anyone running CLI coding agents such as Claude Code, Cursor or Codex fits the Connect flow directly, because the site advertises wiring an agent to the gateway in one command. Solo builders and indie hackers are the second audience. The Free tier carries a monthly allowance, a rotating model lineup and no card requirement, which makes it a low-risk way to test whether DeepSeek V4.1 Flash, MiMo V2.6 Flash or Rudder is good enough before paying anything. Above that, Agent Pass is aimed squarely at scripts and services that call the API around the clock and prefer a flat monthly fee to unpredictable per-call spend. Budget-conscious experimenters get value here too, because the free lineup alone covers three models before any payment is made. Teams with retention or compliance questions are the third. Because the service lets you set retention yourself instead of accepting a baseline policy, and states that paid per-token routes are not kept for training, it appeals to buyers who must document where prompts go. Two caveats matter: this is not a consumer chatbot subscription, and free models follow a separate retention policy that you must opt into. Heavy professional users also fit, paying $99 for Frontier Pass when coding and research workloads outgrow the smaller tiers. Finally, small teams that want one invoice and one usage view instead of a dozen provider dashboards are the fourth group, and the single dashboard pool merging free allowance with pass value is built for exactly that.
Token Harbor Use Cases
Swap one base URL to test Claude, GPT, Gemini, Zhipu and Kimi models without opening new vendor accounts.
Point Claude Code, Cursor or Codex at a single gateway so all agent traffic runs through one key.
Prototype on the free monthly allowance before committing a production workload to a paid pass.
Estimate a month of agent traffic by comparing published request counts per model and pass.
Enforce your own prompt retention window instead of accepting a vendor baseline policy.
Run scripts and background services that call models around the clock on a flat monthly pass.
Serve heavier coding and research work from Frontier Pass when Office Pass usage runs out.
Move between cheap and frontier models by changing the model ID rather than rewriting client code.
Token Harbor Pros and Cons
Pros
- Model integrity is stated as a guarantee: the model ID you send is the model that runs, with no silent swaps.
- A genuine free tier with a monthly allowance, a rotating model lineup and no card required to start.
- Passes combine included usage value, a model boost on selected models and 5% to 15% off pay-as-you-go calls.
- Retention is user-controlled, and paid per-token routes are not kept for training or stored beyond a five-minute exact-match cache.
- One OpenAI-compatible endpoint spans Anthropic, OpenAI, Gemini, Zhipu and Kimi families, plus a one-command connect flow for agents.
Cons
- Pass allowance is usage value spread across four 7-day windows, not a balance: unused room does not carry forward.
- Free models follow a separate retention policy you must opt into, and prompts on them may be kept for diagnostics and product improvement.
- Prices are shown on a monthly view while the model lineup rotates over time, so a given pass quietly changes what it reaches.
FAQ About Token Harbor
Token Harbor Pricing
Freemium: a $0 tier with a monthly free allowance and rotating models, then monthly Passes at $0.99 first month then $1.99 (Agent), $9.99 (Office) and $99 (Frontier) that bundle $10, $35 and $180 of usage value plus 5-15% pay-as-you-go discounts, with a Yearly toggle on the pricing page advertised as the way to save more.
Check official pricingFree
Monthly free allowance with a rotating lineup including DeepSeek V4.1 Flash, MiMo V2.6 Flash and TH-Rudder, promotional models added over time, and one OpenAI-compatible API; no card required.
Agent Pass
$0.99 for the first month, then $1.99. Built for agents and scripts calling the API around the clock: $10 included usage plus the free allowance, model boost up to $20, then 5% off pay-as-you-go.
Office Pass
For desktop agents and everyday writing, research and coding: $35 included usage plus the free allowance, model boost up to $70, then 10% off pay-as-you-go; the most popular pass.
Frontier Pass
Frontier model access for heavy professional work: $180 included usage plus the free allowance, model boost up to $360, then 15% off pay-as-you-go with Claude Opus 5.5 and GPT-6 Astra.
Token Harbor Alternatives
Ecommerce Data API
Amazon product, keyword, traffic and market research data through a REST API, remote MCP server and Agent Skill for developers and AI workflows.
FastRouter
FastRouter is a unified LLM gateway that gives developers one OpenAI-compatible API for 200+ models from providers such as OpenAI, Anthropic, Google, and xAI. It adds automatic model routing, fallbacks, caching, spend caps, role-based access, guardrails, and per-request observability, while passing provider token prices through with no markup on any plan, including the free tier.
MapGO
How far is this point from the sea? From the nearest border? Which country, region and district are it in? MapGO answers those in one click.