AI Agent API

The right model and controllable cost
for every AI agent

From Hermes Agent, OpenClaw and coding agents to support and automation workflows. Use one API for multiple models while managing budgets, usage and Taiwan billing in one place.

Create an Agent API key Choose your agent

Connect your agent with two environment variables

export OPENAI_BASE_URL=https://bazaarlink.ai/api/v1
export OPENAI_API_KEY=sk-bl-...

OpenAI-compatible API, with an additional Anthropic-compatible endpoint for Claude Code.

One entry point for every type of agent

Use an existing tool or place BazaarLink between your custom agent and its models.

Autonomous work and automation

Choose the best balance of speed, quality and cost for scheduled work, research, data processing and long-running agents.

Hermes AgentOpenClawCustom agents

Use cases: Research summaries · Data processing · Scheduled tasks · Multi-step workflows

Coding agents

Keep familiar terminal and IDE workflows while using multiple models through one billing entry point instead of tying the team to one subscription.

ClineKilo CodeOpenCodeCodex CLIClaude Code

Use cases: Coding · Code review · Debugging · Documentation and tests

Support and operations agents

Let agents handle repetitive, low-complexity work that needs 24-hour coverage, with escalation to people when necessary.

Discord BotLINE BotWebsite supportTicketing systems

Use cases: FAQs · Ticket classification · Community management · First drafts

What does BazaarLink do in your agent workflow?

Your agent focuses on completing work. BazaarLink provides model access, cost visibility and billing boundaries, so you do not have to select a new provider for every agent, manage multiple overseas bills or depend on one model.

Your agentBazaarLink APITask-based model selectionUsage and budget

Stay in control after your agent goes live

One compatible API

Use an OpenAI-compatible API as the common entry point without rewriting existing agents and toolchains for every model.

Visible cost for every agent

Give each project or agent an independent key and budget limit to attribute usage and prevent one workflow from running out of control.

Billing for Taiwan teams

TWD top-ups, Taiwan e-invoices and consolidated billing make agent expenses easy to allocate and reimburse.

Start with one real workflow

Not sure what your first agent should do? Begin with community FAQs, first-line support or ticket classification: low-risk, measurable tasks that validate both model and cost configuration.

Use cases for Taiwan teams

Frequently asked questions

Which AI agents does BazaarLink support?

BazaarLink works with custom agents, support systems and automation workflows that use an OpenAI-compatible API, as well as coding agents such as Cline, Kilo Code, OpenCode and Codex CLI. Claude Code can use the separate Anthropic-compatible integration.

Why does an AI agent need a multi-model API?

Tasks have different requirements for speed, quality and cost. Short FAQ answers can use fast, lower-cost models, while complex reasoning, code review or high-risk responses can move to stronger models. This controls cost without sacrificing the experience.

How can I control the budget when several agents run at once?

Create an independent API key and budget for each agent. One agent reaching its limit does not affect other workflows, and billing can be attributed directly to a project or purpose.

Launch one agent, then expand your workflow

One key, traceable usage and clear cost boundaries give your team a safe place to start.

Create an Agent API key

The right model and controllable costfor every AI agent