01 / Getting started

Get your own AI keys

On this page 11

The Synaplan Demo gets a fresh installation chatting with the first message you send, but it is shared, capped at 40 messages a day, and ends after 14 days. Connecting your own AI provider key lifts every Demo limit immediately, gives you full provider throughput, and unlocks the complete model catalog.

Several providers offer free tiers or starter credits, so getting a key takes about two minutes.


Where to paste your key#

  1. Log in as an administrator and open Operate → AI infrastructure → Models & keys (/admin/setup), or click Get your own keys in the Demo banner.
  2. Pick a provider card, paste your API key, and click Save.
  3. Synaplan validates the key live against the provider, stores it AES-256 encrypted in your database, and — if you were on the Synaplan Demo — leaves the Demo automatically and applies the provider's recommended default models.

Changed your mind while testing? Click Undo in the confirmation toast right after saving a provider key to switch back to the Synaplan Demo.

Tip

Prefer environment variables? You can also set GROQ_API_KEY, CEREBRAS_API_KEY, OPENAI_API_KEY, etc. in backend/.env before or after startup. Synaplan imports environment keys into the encrypted store on first use.


Providers, what they unlock, and free-tier availability#

Free-tier and trial-credit summary checked on 2026-10-20. Vendors change their promotions and rate limits over time — always check the provider's pricing page for current terms.

Groq (GROQ_API_KEY)#

  • Where to get a key: console.groq.com/keys
  • Free tier (as of Oct 2026): Free developer tier available after sign-up (no credit card required for basic rate-limited requests); pay-as-you-go removes the tight per-minute caps.
  • What it unlocks: Ultra-fast open-weight chat and sorting (openai/gpt-oss-120b, openai/gpt-oss-20b, qwen/qwen3.6-27b, Llama 3.3 70B), vision (qwen/qwen3.6-27b), and fast audio transcription (whisper-large-v3-turbo).

Cerebras (CEREBRAS_API_KEY)#

  • Where to get a key: cloud.cerebras.ai (API Keys)
  • Free tier / trial (as of Oct 2026): Free tier with rate limits; adding a payment method grants trial credits (typically $5) and unlocks the Developer tier (1,000 RPM / 1M uncached TPM on gpt-oss-120b).
  • What it unlocks: Wafer-scale inference at ~2,000–3,000 tokens/s for gpt-oss-120b (chat, coding, reasoning, memory extraction) and qwen-3.8-27b (chat and vision/OCR).

Google AI Studio — Gemini (GOOGLE_GEMINI_API_KEY)#

  • Where to get a key: aistudio.google.com/apikey
  • Free tier (as of Oct 2026): Google AI Studio offers a free tier with per-minute and daily request quotas on Gemini Flash / Flash-Lite models; billing-enabled projects unlock higher production quotas and media models.
  • What it unlocks: Google Gemini mix in Quick model setup — Gemini 3.1 Pro, Gemini 3.5 / 3 Flash, Gemini 3.1 Flash-Lite (chat + vision), Nano Banana 2 (gemini-3.1-flash-image) and Gemini 3 Pro Image generation, Veo video generation, and Gemini TTS.

Mistral — EU (MISTRAL_API_KEY)#

  • Where to get a key: console.mistral.ai (API Keys)
  • Free tier (as of Oct 2026): "Experiment" free tier available on La Plateforme (phone verification required; data on the free tier may be used for training unless on a billed plan).
  • What it unlocks: Europe mix (cloud option) — French/EU-hosted Mistral Medium 3.5 and Mistral Large 3 (chat + vision), plus Voxtral speech-to-text and text-to-speech.

TrustedTokens — Germany (TRUSTEDTOKENS_API_KEY)#

  • Where to get a key: trustedtokens.eu (Account → API Access)
  • Free tier (as of Oct 2026): Commercial sovereign inference platform (pay-as-you-go; contact TrustedTokens for evaluation credits).
  • What it unlocks: Europe mix (sovereign German GPU hosting in Munich, zero data retention, no CLOUD Act exposure) — GLM 5.2 / 5.3, DeepSeek V4, Qwen 3.6 35B (chat + vision), and GPT OSS 120B.

OpenAI (OPENAI_API_KEY)#

  • Where to get a key: platform.openai.com/api-keys
  • Free tier (as of Oct 2026): Prepaid billing required (typically $5 minimum top-up).
  • What it unlocks: OpenAI GPT mix — GPT-6 Astra / Sol / Luna, GPT-5.6 / 5.5 / 5.4 families (chat + vision + reasoning), gpt-image-2.5-flare and gpt-image-2.5-sunburst (text-to-image and image editing), text-embedding-3-small / large for RAG without local Ollama, Whisper transcription, and OpenAI TTS.

Anthropic (ANTHROPIC_API_KEY)#

  • Where to get a key: console.anthropic.com/settings/keys
  • Free tier (as of Oct 2026): Prepaid billing required (typically $5 minimum top-up).
  • What it unlocks: Anthropic Claude mix — Claude Fable 5.1, Claude Opus 5, Claude Sonnet 5, Claude Opus 4.8, and Claude Haiku 4.5 (chat, coding, vision, and document analysis).

Don't want any cloud key? Run Ollama locally#

If you want every prompt and embedding to stay on your own machine with zero per-token cost, use the local Ollama option instead — see Local AI (Ollama).