Ship Your AI Projects
aamp is a self-hosted, air-gapped AI agent orchestration platform for LLM agents, RAG and workflows.
It runs entirely on your own infrastructure — against local model engines (vLLM, Ollama, LocalAI, TGI) or cloud APIs. You control which model sees which data, what each agent is allowed to touch, and what every call costs, with a full audit trail host-side. Nothing leaves your infrastructure unless you say it can.
- Zero outbound telemetry
- Local or frontier models
- Single Go binary
- Hardware-isolated sandboxing
Build the layer once. Swap models forever.
Where do you want to deploy AI agents?
Engineering, security and operations teams run governed agents with strict host authority, microVM sandboxing and dual safety firewalls. Pick the job you have now.
On your infrastructure. 100% data sovereignty, EU AI Act compliance.
Runs on your infrastructure, in your jurisdiction. No third party to trust. Full network and hardware isolation in a single binary — zero telemetry, no data leaving for the cloud, and hard coverage of GDPR, NIS2 and DORA requirements.
Your own fully isolated AI Control Plane — no data leakage, no telemetry, no regulatory exposure.
Zero PII transfer
- —100% Native Air-Gap — delivers zero PII transfer to external clouds.
- —Dual Safety Firewalls — filter data leaks and Prompt Injection attacks.
Full audit trail
- —Full Audit Trail — 14 log domains written to the local
aamp.sqlite3database. - —Zero Training Risk — your prompts never feed external AI models.
Cyber-resilience
- —Firecracker MicroVM — KVM hardware isolation rules out the blast radius.
- —On-Premise Reliability — processes keep running with no SaaS dependency.
Code isolation
Security
EU Compliance
Predictable AI costs: no token markup and no “Step Tax”.
Pay only for completed business tasks, not for the agent's internal reasoning loops — with full freedom to plug in your own API keys or free local models.
Control your AI budget: 0% token markup and no “Step Tax”.
Run model
Token markup
Access & B2B
Test and compare any model, bound per agent.
Full AI model freedom (Model Agnostic).
Combine any cloud and local models in one system — no vendor lock-in, wholesale rates (BYOK) and zero markup.
Any OpenAI-compatible model, bound per agent — local open weights, hosted frontier, or both.
Combining cloud and local AI (e.g. Ollama)
- —Any Provider: Freely mix OpenAI, Anthropic or Google Bedrock with local open-weight models (Ollama, vLLM).
- —Smart Routing: Send simple queries to cheap local models and complex work to the cloud.
Zero Lock-In & BYOK model
- —Bring Your Own Key: Use your own corporate agreements and pay provider rates directly.
- —Instant Switching: Change model provider in one click, without rebuilding agent logic.
Deterministic Logic (Starlark)
- —No Hallucinated Maths: Combine text synthesis with precise computation in the sandboxed Starlark language.
- —Local PII Protection: Filter sensitive data before a request reaches external model APIs.
Local engines
Cloud providers
Binding and comparison
Four segments running aamp in-house.
One governed platform, four segments running it in-house — each with its own constraint, and the same control layer underneath.
Mid-Market E-commerce & High-Volume Processes
Shops and platforms handling thousands of requests a day (complaints, returns, price updates).
Problem
Runaway costs on Make/Zapier-style platforms, driven by a fee for every step of the decision loop.
What aamp gives them
Execution-Based Billing (0% Step Tax) — a flat fee per completed task instead of a financial penalty for complex agent reasoning.
Software Houses, ISVs and B2B Integrators
B2B software vendors (ERP, CRM, HRM) and automation agencies.
Problem
They need to add AI modules to their own software fast, without risking margin on “Step Tax” and tokens.
What aamp gives them
Embeddable AI Engine — plug a single Go binary in over REST API and ship your own AI features 10x faster under your own brand (White-Label / OEM).
Regulated Sector (Healthcare, Healthtech, Finance, Legal)
Organisations processing sensitive data (PII, PHI, transactional records) under GDPR, NIS2, DORA and financial supervision.
Problem
Sending data to public clouds is off the table because of leak and compliance risk.
What aamp gives them
100% Air-Gap & MicroVM — the agent runs scripts and analyses documents in a hardware-isolated environment using local models (Ollama/vLLM).
Manufacturing, Logistics and Legacy Systems (ERP / WMS / CRM)
Mature SME and mid-market companies on traditional infrastructure (e.g. Comarch XL, SAP On-Prem).
Problem
No modern APIs, and scraping supplier or government portals risks infecting the core database.
What aamp gives them
Headless Browser Sandboxing — safe browser automation and file reading inside a disposable MicroVM.
One console for every control plane.
Use aamp when…
Four operational signals that your organisation has outgrown simple SaaS tools and needs a production Control Plane.
When SaaS bills for AI decision loops spiral out of control
Your workflows in Make, Zapier or n8n rack up huge costs through the “Step Tax” — a fee for every step and every retry inside the agent’s reasoning loop.
Execution-Based Billing — you pay only for 1 completed business run, with no markup on LLM tokens (BYOK).
When a compliance audit (EU AI Act / GDPR / NIS2 / DORA) blocks the rollout
Legal or the DPO rejects cloud AI tools over PII leak risk, a missing audit trail or US CLOUD Act exposure.
100% Native Air-Gap — deployment with 14 log domains recorded in the local aamp.sqlite3 database and full network isolation.
When unknown code and browser scripts threaten your ERP/CRM databases
Agents have to log into supplier portals, download PDF/Excel attachments or run scripts, which creates Prompt Injection and infrastructure infection risk.
Firecracker MicroVM Sandboxing — every untrusted file and browser session runs in a disposable, hardware-isolated KVM microVM with zero blast radius.
When a clustered stack (Docker/Postgres/Redis) paralyses maintenance
DevOps loses time maintaining and upgrading multi-container environments (e.g. Dify.ai / Langflow) in on-premise deployments.
Single Go Binary Architecture — the whole platform with all its components starts from one binary in seconds, keeping maintenance simple.
Automation platforms connect apps. aamp governs agents.
The orchestration layer is its own category. Here is where aamp sits in it.
Swipe horizontally
| Criterion / Capability | aamp | Cloud SaaS (e.g. Lindy, Make) | OS Frameworks (e.g. Dify, Flowise) | Scripts / Workflow (e.g. n8n) |
|---|---|---|---|---|
| Deployment model | Single Go Binary + SQLite | Pure SaaS Cloud | Docker / Cloud | Node.js / Docker |
| Sovereignty and Air-Gap | 100% Native (Zero Telemetry) | None (data in the cloud) | Partial (needs configuration) | Partial |
| Code execution isolation | Firecracker MicroVM (Hardware) | None (shared environment) | None (standard container) | None (container / Node) |
| LLM security | Dual Safety Firewalls (Prompt & PII) | Basic | Basic / none | None (needs external services) |
| Billing model | 0% Step Tax + BYOK | Credit packs / Step Tax | Message or query count | Per workflow execution |
Try aamp in the sandbox.
Free sandbox · Self-host in minutes · Talk to us
Partner with aamp.
Our partners offer a range of professional services — expertise and technology capabilities to help our customers both make the first step and enrich their solutions.
Become a Partner →