Can I use my own API keys or a local model with PostMimic?

Yes—cloud AI, own keys (OpenAI/Anthropic/DeepSeek/xAI), or local Ollama; same voice profile + publish workflow. Model choice still sits under brand-safe rules before Agent Mode. You choose how generation is powered without rebuilding how you draft, review, schedule, and publish.

Three power options

PostMimic is built so model choice is a setting, not a separate product:

1. Cloud AI (managed) — Use the platform’s hosted path when you want the simplest setup: no key inventory, no local GPU, generate and go.
2. Bring your own keys (BYOK) — Plug in API keys from providers you already pay: OpenAI, Anthropic, DeepSeek, xAI, and others supported in-product. Usage bills to your provider account; PostMimic still owns voice, rules, and publish.
3. Local via Ollama — Run compatible models on your machine through Ollama when you want generation to stay on hardware you control.

All three feed the same loop: writing-voice profile → draft for a platform → human review → schedule/publish. Voice here means writing voice from real posts—not audio deepfakes.

You can work on phone, web, or desktop independently; model routing depends on what each surface supports for keys and local runtimes, but the voice profile and approval habits stay consistent.

When BYOK makes sense

Bring your own keys when:

• You already have enterprise agreements or credits with OpenAI, Anthropic, DeepSeek, or xAI
• Finance wants model spend on a vendor invoice you already negotiate
• You prefer a specific model family for tone or latency
• Compliance wants a clear line between “app” and “model provider”

Pricing note: on paid PostMimic plans, the BYOK discount of $10/mo is live—check current pricing on the site for plan details. You still get the voice + rules + publish workflow; you supply the generation keys.

BYOK is not “raw ChatGPT in a different skin.” The point of PostMimic remains history-trained drafts and a publish path with optional brand-safe rules—not a blank chat box.

When Ollama (local) makes sense

Local Ollama fits when:

• You want drafts generated on a machine you administer
• You are experimenting with open models before committing cloud spend
• Network or policy constraints push you toward on-device inference
• Developers on the team already run Ollama for other tools

Tradeoffs are the usual ones: you manage hardware, model updates, and performance. The product expectation stays the same—drafts should still go through human review before they represent the brand publicly.

Local generation does not remove the need for a good training archive. Garbage-in still produces mush; a strong posting history still produces recognizable voice.

What stays the same across cloud, BYOK, and Ollama

Regardless of power source:

• Writing-voice profile built from real posts
• Platform-aware drafts (LinkedIn vs X vs others)
• Brand-safe rules when you configure them (topics, forbidden words, alliances/opponents where relevant)
• Human approve before publish (and before trusting any Agent Mode)
• WordPress Connect category matching to exact existing terms (or Uncategorized)—the plugin does not create categories

You should not rebuild voice every time you switch keys. Swap the engine; keep the fingerprint.

Human review still matters

Model choice changes cost and control. It does not replace judgment:

• Approve or edit drafts before they go live
• Keep Agent Mode off until voice + rules feel stable under supervision
• Separate client or candidate voices so keys never mix profiles
• Watch for provider outages or local resource limits the same way you watch any dependency

AI drafts; humans own the account.

Switching paths without losing voice

Teams often start on managed cloud, then move BYOK when invoices matter, or try Ollama for sensitive drafts. Do this cleanly:

1. Confirm the voice profile still points at the same training set
2. Generate a small batch on the new path and diff against recent approved posts
3. Keep brand rules enabled across the switch
4. Do not enable Agent Mode in the same week you change engines

The goal is interchangeable power sources behind one workflow—not three disconnected experiments.

Try Free

Curious whether managed cloud, BYOK, or local fits your team? Start free at https://postmimic.app, build a voice from real posts, generate with the path you prefer when available, and keep the same review-before-publish habit and the brand-safe rules before Agent Mode checklist. Paid BYOK plans include the live $10/mo discount—see pricing for current plan math.