Feature Request: Bring-Your-Own-LLM Key Mode connect your own OpenAI, Anthropic, DeepSeek, Minimax, Kimi, Gemini API key to power TryPost's AI features

[githb-ac]

TryPost's AI copilot is one of its most powerful differentiators, but for high-volume users, agencies, and enterprise accounts the cost of AI-generated content is a meaningful consideration particularly when AI usage is metered into pricing tiers or capped at a fixed number of generations per month. This feature request proposes a Bring-Your-Own-LLM Key Mode that allows users to connect their own API key from OpenAI, Anthropic, Google Gemini, or other supported providers, routing all of TryPost's AI features through the user's own account and billing relationship rather than TryPost's shared infrastructure giving power users unlimited AI generation at their own cost while opening a clear enterprise pricing path for TryPost.

Summary

A workspace-level setting where users can input their own API key from a supported LLM provider. When a key is configured, all AI copilot features post drafting, caption rewriting, hashtag suggestions, content repurposing, sentiment analysis, and competitor digests run through the user's own API account rather than TryPost's. The experience for the user is identical; the only difference is that AI generation costs are billed directly to their provider account rather than counted against a TryPost usage tier.

Why This Matters

Bring-Your-Own-Key is a standard enterprise and power-user feature in AI-native tools and signals a level of platform maturity and transparency that builds trust with technical buyers. For agencies running AI generation across dozens of client workspaces, or for individual creators who use the AI copilot heavily every day, hitting a generation cap or paying a premium AI tier is a friction point that can drive churn. BYOK mode removes that ceiling entirely while simultaneously reducing TryPost's own LLM infrastructure costs for its highest-usage accounts a rare feature that benefits both the user and the platform. It also allows users to select their preferred model, meaning teams that have standardised on Claude for brand voice reasons or GPT-4o for output quality can use their preferred model natively inside TryPost without compromise.

Proposed MVP

API key input in workspace settings supporting OpenAI, Anthropic Claude, DeepSeek, Minimax, Kimi and Google Gemini as initial providers

Key validation on entry with a clear success or error state so users know immediately whether their key is working

Model selector allowing users to choose which specific model version to use for example GPT-4o, GPT-4o mini, Claude Sonnet, or Gemini Pro with a note on the capability and cost tradeoffs of each

All existing AI copilot features route through the user-supplied key when one is configured with no degradation in functionality or UX

Usage transparency panel showing estimated token consumption per session and a link to the user's provider dashboard for billing details

Fallback option allowing users to define whether TryPost should fall back to its own shared AI infrastructure if the user's key hits a rate limit or error, or whether AI features should be disabled entirely until the issue is resolved

Clear indication in the UI when BYOK mode is active so users always know which AI infrastructure is powering their session

Key stored securely with encryption at rest and never exposed in plaintext after initial entry

[richardchen874-sys]

Great feature request! Bring-Your-Own-LLM Key Mode makes a lot of sense for high-volume users, agencies, and enterprise accounts.

This is exactly why we started AI Token Pro - to help teams reduce their AI API costs. We offer discounted US-region AI tokens that can help you reduce costs by 30-50%, and we work with many agencies and enterprise customers who see significant savings.

Would love to know more about your AI usage - how many generations per month are you anticipating?

[paulocastellano]

@githb-ac @richardchen874-sys you already can do it in self-hosted mode.

For Cloud version i don't have any intention to support it, it's super high risky host 3th party LLM's tokens.

[richardchen874-sys]

That makes sense, and I agree with the risk concern. Storing arbitrary third-party LLM tokens for Cloud users creates a lot of security, support, and liability surface area.

A safer pattern might be to keep BYOK only for self-hosted, and for Cloud avoid storing customer provider keys entirely. If Cloud AI cost becomes a bottleneck, another option is to test a managed OpenAI-compatible endpoint with strict usage caps and usage receipts, so TryPost controls the provider relationship instead of holding each user's OpenAI/Anthropic tokens.

For your Cloud version, is the bigger concern today LLM infrastructure cost, token custody risk, or support complexity when users bring different providers? I am testing an OpenAI-compatible API layer focused on official Chinese models, and if cost comparison ever becomes useful, I can share a small private-beta credit for a limited workflow test.