各个ai工具的自定义api使用

星智通网关API接入指南一、个人key获取步骤登录FAI平台 无此平台账号可联系郭若鸿开通)二、接口验证curl https://aigw.finloopai.ai/v1/messages \ -H x-api-key: $YOUR_API_KEY \ -H anthropic-version: 2023-06-01 \ -H Content-Type: application/json \ -d { model: claude-opus-4-6, max_tokens: 1024, messages: [ { role: user, content: Explain quantum entanglement in one paragraph. } ] }三、Claude Code 接入方式1. 直接编辑配置文件配置文件位于~/.claude/settings.json全局或项目根目录.claude/settings.json项目级mac和linux可添加以下配置{ apiBaseUrl: https://aigw.finloopai.ai, apiKey: ak-xxxxxxxxxxxxxxxx }WINDOWS 可在配置文件下加以下配置模型和KEY可对应替换env: { ANTHROPIC_API_KEY: ak-xxx, ANTHROPIC_AUTH_TOKEN: ak-xxx, ANTHROPIC_BASE_URL: https://aigw.finloopai.ai, ANTHROPIC_DEFAULT_HAIKU_MODEL: claude-sonnet-4.6, ANTHROPIC_DEFAULT_OPUS_MODEL: claude-opus-4.7, ANTHROPIC_DEFAULT_SONNET_MODEL: claude-opus-4.6, ANTHROPIC_MODEL: claude-sonnet-4.6, ANTHROPIC_REASONING_MODEL: claude-sonnet-4.6, CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS: 1 }方式2. CC SwitchDesktop 可视化方式推荐新手启动CC Switch点击Codex选项卡点击新增配置在配置界面中配置网关url网关API key和网关提供的模型然后点击添加保存配置打开 Claude Code Desktop 会自动跳过登录如果已登录需要先退出登录并重启 Claude Code Desktop测试对话确认是否正常使用四、Codex 接入方式启动CC Switch点击Codex选项卡点击新增配置在配置界面中配置网关url网关API key和网关提供的模型然后点击添加保存配置打开Codex如果已登录先退出登录选择其他方式登录然后输入网关API key对话确认是否正常使用备注所有配置改动需要杀掉codex进程重新打开应用才会生效。五、OpenCode接入方式方式一直接修改配置文件OpenCode的配置文件默认路径在~/.config/opencode/opencode.json将以下配置添加到配置文件中{ $schema: https://opencode.ai/config.json, disabled_providers: [], provider: { Finloop-gateway: { name: Finloop-gateway, npm: ai-sdk/openai-compatible, options: { baseURL: https://aigw.finloopai.ai/v1, apiKey: {your-api-key} }, models: { claude-opus-4.6: { name: claude-opus-4.6 }, claude-opus-4.7: { name: claude-opus-4.7 }, claude-sonnet-4.6: { name: claude-sonnet-4.6 } } } } }重启 OpenCode 即可方式二使用 GUI 配置界面进入 OpenCode 桌面端点击切换模型选择自定义供应商测试对话确认是否正常使用六、Cursor 接入方式准备工作(1)下载安装好 Cursor-BYOK 助手下载地址https://github.com/leookun/cursor-byok/releases选择自己电脑对应的软件包下载并点击下载好的文件进行安装(2)打开cursor进行配置(i) 点击Cursor设置按钮并进入models界面检查一下这里不要勾选(ii) 确认无误后点击Network tap进入网络配置界面开启HTTP/1.1必须开启(iii) ***先登出Cursor公司账户***关闭cursor继续下一部分的配置配置接入模型 以cluade-opus-4.6为例(1) 打开Cursor - BYOK 助手启动助手检查状态是服务运行中并确保下面第一个红框的按钮没有开启(2) 模型配置省事优先方式一快速配置**开发测试、产品、设计模型******以下提供了三份标准config模板可供对应同事快速应用不用手动一个个配若有其他需要的模型自己补充即可开发config.developer.yamllog: falseproviderStreamIdleTimeout: 240backendListenAddr: 127.0.0.1:18090proxyListenAddr: 127.0.0.1:18080modelAdapters:displayName: claude-opus-4.8type: anthropicbaseURL: https://aigw.finloopai.ai/v1/messagesapiKey: REPLACE_WITH_API_KEYtooltipData: 备注modelID: claude-opus-4-8reasoningEffort: mediumopenAIEndpoint: “”openAIExtraParamsEnabled: falseopenAIExtraParamsJSON: “”customHeadersEnabled: falsecustomHeadersJSON: “{}”anthropicExtraParamsEnabled: falseanthropicExtraParamsJSON: |-{}contextWindowTokens: 0maxCompletionTokens: 0anthropicMaxTokens: 0anthropicThinkingEffort: xhighthinkingBudgetTokens: 0displayName: claude-fable-5type: anthropicbaseURL: https://aigw.finloopai.ai/v1/messagesapiKey: REPLACE_WITH_API_KEYtooltipData: 备注modelID: claude-fable-5reasoningEffort: mediumopenAIEndpoint: “”openAIExtraParamsEnabled: falseopenAIExtraParamsJSON: “”customHeadersEnabled: falsecustomHeadersJSON: “{}”anthropicExtraParamsEnabled: falseanthropicExtraParamsJSON: |-{}contextWindowTokens: 0maxCompletionTokens: 0anthropicMaxTokens: 0anthropicThinkingEffort: xhighthinkingBudgetTokens: 0displayName: claude-sonnet-5type: anthropicbaseURL: https://aigw.finloopai.ai/v1/messagesapiKey: REPLACE_WITH_API_KEYtooltipData: 备注modelID: claude-sonnet-5reasoningEffort: mediumopenAIEndpoint: “”openAIExtraParamsEnabled: falseopenAIExtraParamsJSON: “”customHeadersEnabled: falsecustomHeadersJSON: “{}”anthropicExtraParamsEnabled: falseanthropicExtraParamsJSON: |-{}contextWindowTokens: 0maxCompletionTokens: 0anthropicMaxTokens: 0anthropicThinkingEffort: xhighthinkingBudgetTokens: 0displayName: kimi-k3type: openaibaseURL: https://aigw.finloopai.ai/v1apiKey: REPLACE_WITH_API_KEYtooltipData: 备注modelID: kimi-k3reasoningEffort: xhighopenAIEndpoint: /v1/chat/completionsopenAIExtraParamsEnabled: falseopenAIExtraParamsJSON: |-{“service_tier”: “priority”}customHeadersEnabled: falsecustomHeadersJSON: “{}”anthropicExtraParamsEnabled: falseanthropicExtraParamsJSON: “”contextWindowTokens: 0maxCompletionTokens: 0anthropicMaxTokens: 0thinkingBudgetTokens: 0displayName: gpt-5.6-soltype: openaibaseURL: https://aigw.finloopai.ai/v1apiKey: REPLACE_WITH_API_KEYtooltipData: 备注modelID: gpt-5.6-solreasoningEffort: xhighopenAIEndpoint: /v1/chat/completionsopenAIExtraParamsEnabled: falseopenAIExtraParamsJSON: |-{“service_tier”: “priority”}customHeadersEnabled: falsecustomHeadersJSON: “{}”anthropicExtraParamsEnabled: falseanthropicExtraParamsJSON: “”contextWindowTokens: 0maxCompletionTokens: 0anthropicMaxTokens: 0thinkingBudgetTokens: 0displayName: gpt-5.6-terratype: openaibaseURL: https://aigw.finloopai.ai/v1apiKey: REPLACE_WITH_API_KEYtooltipData: 备注modelID: gpt-5.6-terrareasoningEffort: xhighopenAIEndpoint: /v1/chat/completionsopenAIExtraParamsEnabled: falseopenAIExtraParamsJSON: |-{“service_tier”: “priority”}customHeadersEnabled: falsecustomHeadersJSON: “{}”anthropicExtraParamsEnabled: falseanthropicExtraParamsJSON: “”contextWindowTokens: 0maxCompletionTokens: 0anthropicMaxTokens: 0thinkingBudgetTokens: 0displayName: gemini-3.1-flash-lite-imagetype: openaibaseURL: https://aigw.finloopai.ai/v1apiKey: REPLACE_WITH_API_KEYtooltipData: 备注modelID: gemini-3.1-flash-lite-imagereasoningEffort: xhighopenAIEndpoint: /v1/chat/completionsopenAIExtraParamsEnabled: falseopenAIExtraParamsJSON: |-{“service_tier”: “priority”}customHeadersEnabled: falsecustomHeadersJSON: “{}”anthropicExtraParamsEnabled: falseanthropicExtraParamsJSON: “”contextWindowTokens: 0maxCompletionTokens: 0anthropicMaxTokens: 0thinkingBudgetTokens: 0displayName: grok-4.5type: openaibaseURL: https://aigw.finloopai.ai/v1apiKey: REPLACE_WITH_API_KEYtooltipData: 备注modelID: grok-4.5reasoningEffort: xhighopenAIEndpoint: /v1/chat/completionsopenAIExtraParamsEnabled: falseopenAIExtraParamsJSON: |-{“service_tier”: “priority”}customHeadersEnabled: falsecustomHeadersJSON: “{}”anthropicExtraParamsEnabled: falseanthropicExtraParamsJSON: “”contextWindowTokens: 0maxCompletionTokens: 0anthropicMaxTokens: 0thinkingBudgetTokens: 0displayName: deepseek-v4-protype: openaibaseURL: https://aigw.finloopai.ai/v1apiKey: REPLACE_WITH_API_KEYtooltipData: 备注modelID: deepseek-v4-proreasoningEffort: xhighopenAIEndpoint: /v1/chat/completionsopenAIExtraParamsEnabled: falseopenAIExtraParamsJSON: |-{“service_tier”: “priority”}customHeadersEnabled: falsecustomHeadersJSON: “{}”anthropicExtraParamsEnabled: falseanthropicExtraParamsJSON: “”contextWindowTokens: 0maxCompletionTokens: 0anthropicMaxTokens: 0thinkingBudgetTokens: 0displayName: deepseek-v4-flashtype: openaibaseURL: https://aigw.finloopai.ai/v1apiKey: REPLACE_WITH_API_KEYtooltipData: 高性价比编码、推理与日常开发任务modelID: deepseek-v4-flashreasoningEffort: xhighopenAIEndpoint: /v1/chat/completionsopenAIExtraParamsEnabled: falseopenAIExtraParamsJSON: |-{“service_tier”: “priority”}customHeadersEnabled: falsecustomHeadersJSON: “{}”anthropicExtraParamsEnabled: falseanthropicExtraParamsJSON: “”contextWindowTokens: 0maxCompletionTokens: 0anthropicMaxTokens: 0thinkingBudgetTokens: 0displayName: glm-5.2type: openaibaseURL: https://aigw.finloopai.ai/v1apiKey: REPLACE_WITH_API_KEYtooltipData: 长任务执行、代码工程与复杂工具调用modelID: glm-5.2reasoningEffort: xhighopenAIEndpoint: /v1/chat/completionsopenAIExtraParamsEnabled: falseopenAIExtraParamsJSON: |-{“service_tier”: “priority”}customHeadersEnabled: falsecustomHeadersJSON: “{}”anthropicExtraParamsEnabled: falseanthropicExtraParamsJSON: “”contextWindowTokens: 0maxCompletionTokens: 0anthropicMaxTokens: 0thinkingBudgetTokens: 0displayName: qwen3.7-maxtype: openaibaseURL: https://aigw.finloopai.ai/v1apiKey: REPLACE_WITH_API_KEYtooltipData: 新版 Qwen 综合模型适合编码、中文任务与通用推理modelID: qwen3.7-maxreasoningEffort: xhighopenAIEndpoint: /v1/chat/completionsopenAIExtraParamsEnabled: falseopenAIExtraParamsJSON: |-{“service_tier”: “priority”}customHeadersEnabled: falsecustomHeadersJSON: “{}”anthropicExtraParamsEnabled: falseanthropicExtraParamsJSON: “”contextWindowTokens: 0maxCompletionTokens: 0anthropicMaxTokens: 0thinkingBudgetTokens: 0routing:mode: localhomeMetrics:includeCacheWriteInHitRate: falselastAgentModelHash: “”在这里插入代码片产品config.product.yaml设计config.design.yamlWindowsMAC同理Step0 下载保存后将config中的api key 替换成自己的ctrl H 可快速替换Mac :Command Option FStep1 点击”设置文件夹“Step2 **重点先关闭再复制****先关闭Cursor助手**再将刚刚配了你的key的config复制到这里。记得改名回configStep3 重新打开Cursor助手Step4点击”模型配置“可以看见已经配好了所需的模型。**方式二**手动配置适合自定义所需要的模型点击第二个红框中的模型配置按钮进入配置页面点击保存并测试如果下面返回了“xxx t/s|首字 xx.x s”即为链接成功正式使用(1) 打开 Cursor刷新模型列表(2) 回到 Curosr Agent对话窗口选择对应模型进行测试可以看到输出的模型名称和我们配置的一致这样就成功在cursor中用自己的url和key配置对应模型啦注意原公司账号的对话无法继续使用需要重新开启对话或者重新创建workspace七、Hermes Agent 接入自有模型Hermes Agent 是 NousResearch 出品的终端 AI Agent 工具支持所有 OpenAI-compatible 端点。方式1. 交互式配置推荐hermes model# 选择 Custom endpoint# 按提示输入Base URL、API Key、模型名、Context Window方式2. 直接编辑配置文件配置文件位于 ~/.hermes/config.yamlAPI Key 存入 ~/.hermes/.env~/.hermes/.envDELEGATION_KEYyour-api-key~/.hermes/config.yamlmodel:provider: customdefault: your-model-namedelegation:base_url: https://your-gateway/v1api_key: ${DELEGATION_KEY}model: your-model-name方式3. CLI 快速设置hermes config set model your-model-namehermes config set DELEGATION_KEY your-api-key# 自动写入 .env方式4. 直接和Hermes Agent对话进行设置八、JetBrains IDEA AI Assistant 接入支持 IntelliJ IDEA、PyCharm、GoLand 等全系 JetBrains 产品三平台配置路径一致。需要 2024.1 及以上版本。打开SettingsmacOS⌘,Windows / LinuxCtrlAltS导航到Tools → AI Assistant → Custom Models点击新增填写字段值NameFinloop-claude-opus自定义Base URLhttps://aigw.finloopai.ai/v1API Keyak-xxxxxxxxxxxxxxxxModel IDclaude-opus-4-6点击Test Connection验证在 AI Assistant 对话框的模型下拉列表中选择该模型