Deploy AI in minutes.
Your cloud. Your model. Your region.
Pick an open-weight model and a region, sign in with AWS, and forgestack generates and deploys a complete stack — model host, API gateway, knowledge base and fine-tuning — entirely inside your own cloud account. No ML platform team required, whatever size your org is.
- open models
- 11
- AWS regions
- 30
- provisioning
- ~30s
- to prototype
- $0
The pipeline
From model to live API in minutes, not sprints
Pick a model & region
Choose an open-weight model from the catalog, and where it runs — 30 AWS regions, sized to the model automatically.
- Open-weight, Hugging Face-hosted catalog
- 30 AWS regions, instance sized automatically
- Swap models anytime — API contract never changes
Connect your cloud
Pick whichever authentication your account supports — SSO, a cross-account role, access keys, or a local profile. No credential ever touches our servers unencrypted.
- 4 auth paths — SSO to plain access keys
- sts:AssumeRole verified live, not assumed
- Zero long-lived credentials touch our servers
We forge the template
Your config becomes a real, prefilled CloudFormation template — inference host, API gateway, knowledge base. Downloadable, owned by you.
- Real CloudFormation, fully prefilled & readable
- VPC, IAM role, API gateway all defined
- Downloadable — the file is yours, not ours
Apply — we provision
Under your assumed role, we stand up a VPC, a dedicated inference node, an encrypted S3 knowledge base and an API gateway, then warm the model — with live per-resource progress.
- VPC + subnets stood up under your role
- GPU/CPU node launched, model weights pulled
- Encrypted S3 knowledge base + API gateway live
Live: chat, cite, fine-tune
Grounded chat with source citations, a documents API backed by your own bucket, and LoRA fine-tuning on your own examples — all in your account.
- Grounded chat, source citations on every reply
- Documents API — reads and writes your own bucket
- LoRA fine-tune on your own question/answer pairs
What we deploy for you
A real stack, not a chatbot wrapper
Every forgestack solution is a self-contained architecture: your knowledge lives in your storage, your model runs on a dedicated compute node in the region you chose, and an API gateway exposes it all behind one key.
Private by design
Your data stays inside your stack — your bucket, your model host, your keys. Nothing is shared across tenants.
Real cloud deployment
AWS role-assumption and stack apply happen against your account via STS + CloudFormation — auditable end to end.
Template you own
Every deployment is a downloadable, deployable template. Version it, audit it, run it yourself.
Model catalog
Pick a model, we host it in your region
Free-tier open-weight models, sized to an instance automatically. Swap models anytime — the API contract never changes.
Qwen2.5 0.5B Instruct
Qwen/Qwen2.5-0.5B-Instruct
Best answers that still fit a free t3.micro. Chat + RAG, no GPU.
SmolLM2 360M Instruct
HuggingFaceTB/SmolLM2-360M-Instruct
Half the weights of Qwen 0.5B. Fastest replies on a 2 vCPU box.
SmolLM2 135M Instruct
HuggingFaceTB/SmolLM2-135M-Instruct
Smallest that still holds a conversation. Good for narrow RAG bots.
TinyLlama 1.1B Chat
TinyLlama/TinyLlama-1.1B-Chat-v1.0
Featherweight. Runs on CPU-only nodes, sub-second replies.
Phi-2
microsoft/phi-2
Microsoft's compact reasoning model. Great for FAQ + support bots.
Qwen 2.5 3B Instruct
Qwen/Qwen2.5-3B-Instruct
Long context in a small footprint. Strong multilingual support.
Zephyr 7B β
HuggingFaceH4/zephyr-7b-beta
DPO-aligned chat model with a natural conversational tone.
Mistral 7B Instruct v0.2
mistralai/Mistral-7B-Instruct-v0.2
The general-purpose default — long documents, multi-turn support, RAG answers.
SmolLM2 1.7B Instruct
HuggingFaceTB/SmolLM2-1.7B-Instruct
Tiny but instruction-tuned properly. Good CPU-node default for short replies.
Phi-3 Mini 4K Instruct
microsoft/Phi-3-mini-4k-instruct
Punches far above its size on reasoning and instruction following.
Qwen 2.5 7B Instruct
Qwen/Qwen2.5-7B-Instruct
Strongest all-rounder here — coding, maths, 29 languages, 32K context.
Need a specific open-weight model? Ask and we'll add it to the catalog.
Access control
Keys scoped to exactly what they can do
Issue a key per environment, per team, or per vendor — and pick which of the three endpoints it can call. Narrow the scopes below to see the key that comes out the other side.
Revoke or re-scope any key instantly — nothing else changes.
The solution API
One key. Three endpoints.
Infinite integrations.
The moment your stack goes live, every endpoint answers — from inside your own account. Feed the knowledge base and kick off fine-tunes through the dashboard or the API, and read the full reference anytime at /docs.
- Per-stack API keys
Issue and revoke keys per environment, per team, per vendor.
- OpenAI-compatible shape
chat.completion responses your existing tooling already understands.
- Grounded answers
Retrieval over your knowledge base with source citations on every completion.
- Documents endpoint
Add, list and remove knowledge-base documents — stored in your own S3 bucket.
- Fine-tune on your data
Kick off a real LoRA fine-tune from your own question/answer pairs, on your own node.
curl -X POST https://your-forgestack.app/api/{projectId}/chat \
-H "x-api-key: fsk_9f2c..." \
-H "content-type: application/json" \
-d '{"messages":[{"role":"user","content":"What is your refund policy?"}]}'Integrations
Built on infrastructure you already trust
AWS
EC2, S3, Lambda, API Gateway, IAM — real CloudFormation under your role.
Hugging Face Hub
Open-weight model weights pulled straight from the Hub at deploy time.
CloudFormation
Every stack is a real, downloadable, re-runnable template you own.
GCP
Compute Engine + GCS — emulated locally via fake-gcs-server (port 4443)
Stop wiring chatbots.
Start forging stacks.
Pick a model and a region, connect your cloud, and get a production-grade, grounded API with a deployable infrastructure template — deployed entirely in your own account, in minutes. Built to make enterprise-grade AI infrastructure simple enough for a team of any size.