Engineer AI.
Build Intelligence.
A local-first AI engineering platform with built-in RAG, MCP tools, intelligent agents, a private AI Gateway and governed workspaces — everything you need to prototype with AI today and ship production automation tomorrow.
12 customers with citations from `customer-ledger.pdf` · query 4ms
What is AI Forge?
The AI engineering platform your stack has been missing
MageTech AI Forge is a self-hosted platform that brings knowledge, agents, tools and model management into a single governed workspace — replacing the glue code you would otherwise build yourself.
One platform, many modules
Chat, knowledge bases, RAG, agents, MCP tools, AI Gateway and AgentOps live in one system — data moves between them without duct tape.
Local-first by default
The entire AI stack can run on your own hardware with Ollama — generation and embeddings stay inside your perimeter.
Governed workspaces
Tenant isolation, role-based access, tool permissions and audit trails make AI features meet the same standard as your other systems.
Standards, not silos
Open APIs, SSE streaming and MCP (Model Context Protocol) keep the platform composable and future-proof.
Honest about readiness
Every capability is labeled Available, Planned or Future — so you always know what ships today versus what is on the roadmap.
OpenAPI / SDK-ready
Generate clients straight from the OpenAPI schema, and stream agent + chat responses via REST and SSE from day one.
Honest framing. This document describes the product and its roadmap at a high level, based on the public repository. Capabilities and plans are labeled as such; nothing is overclaimed or presented as production-ready unless it is.
Platform
Everything you need to build with AI — in one workspace
Eight integrated modules. Each one is independently useful; together they cover the full AI engineering lifecycle.
Multi-turn chat, grounded
Stream responses over SSE with cited sources from your knowledge base. Conversations stay in your workspace — nothing leaves your perimeter.
Model switching
Route chats across local Ollama models or future managed providers through the AI Gateway, per conversation.
Upload & parse
PDF, DOCX, TXT and Markdown. Files are parsed, cleaned, chunked and enriched with metadata before they are indexed.
One source of truth
Knowledge bases are workspace-scoped, versioned and searchable — your AI answers from what you actually know.
Semantic retrieval
Chunked content is embedded locally (nomic-embed-text, 384-d, cosine similarity) and retrieved with source citations.
Today & roadmap
Vector store runs in-process with your data today; native pgvector integration is planned for scale.
Agent runtime
Agents plan, call tools and compose grounded answers. Every execution is recorded with inputs, steps and tool calls.
Orchestration
LangGraph workflows are on the roadmap for deterministic, testable multi-step automation.
Tightly guarded tools
MCP servers & tools over SSE JSON-RPC, with built-in tools (memory, calculator, time) and workspace-level enable/disable.
Safety by design
No shell, OS or arbitrary code execution via AI tools; high-risk and external actions are audited.
One client, many models
The AI Gateway abstracts providers behind a single client (`LLMClient`) with `auto`, `ollama` and `mock` modes.
Future providers
OpenAI, Gemini and Anthropic connectors are planned — keeping local models the default while opening the door to cloud.
Executions & usage
Agent executions, tool calls and usage analytics are recorded per workspace — a window into how AI is actually used.
Advanced AgentOps
Deeper evaluation, testing and monitoring tooling is on the roadmap for mission-critical deployments.
Pipelines
How the pieces fit together
Three animated flows show how a document becomes a grounded answer, how an agent reaches a business system, and how tools are safely exposed.
RAG pipeline Available
In-process vector store today · native pgvector planned.
Agent pipeline Available
LangGraph orchestration planned.
MCP flow Available
SSE JSON-RPC · no shell / arbitrary code.
Features
Capabilities with clear readiness
Every feature is labeled Available, Planned or Future — the honest picture of what ships today versus what is coming.
Multi-turn chat with SSE streaming and citations from your knowledge base.
Upload PDF, DOCX, TXT and Markdown; parse, chunk and index per workspace.
Ground answers in your documents with source citations on every response.
Agent runtime with recorded executions — inputs, steps and tool calls.
MCP servers & tools over SSE with built-in memory, calculator and time tools.
Single client (`LLMClient`) with auto, ollama and mock provider modes.
Tenant-scoped knowledge, agents, tools and models via `X-Workspace-Id`.
JWT access (15 min) + rotating HttpOnly refresh tokens, Argon2id password hashing.
Token-bucket per client: auth 10/min, default 200/min, Redis with in-memory fallback.
Executions, tool calls and usage recorded per workspace for observability.
Central catalog with local models and metadata; multi-provider planned.
Self-host via Docker Compose — Postgres 16, Redis 7 and Ollama in your estate.
Role model with rank ordering defined; route-level enforcement is next milestone.
Vector search directly in Postgres with pgvector for consistent storage.
Deterministic, testable multi-step agent workflows.
Revoke compromised sessions server-side.
Trigger external workflows from platform events.
Enterprise single sign-on and delegated authorization.
OpenAI, Gemini and Anthropic connectors at the AI Gateway.
Deeper integration with popular agentic frameworks.
CRM, ERP and helpdesk integrations are being explored.
Architecture
How it is built
A pragmatic, self-hostable stack — proven, boring technology where it matters, modern where it counts.
Backend
Python · FastAPI · Pydantic · SQLAlchemy · Alembic. OpenAPI schema at /docs.
Frontend
Next.js 16 · React 19 · TypeScript · Tailwind CSS 4 — a modern, fast UI.
AI
Ollama runtime (`qwen2.5:7b` for generation, `nomic-embed-text` for embeddings) via the AI Gateway.
Persistence
PostgreSQL 16 (primary store) + Redis 7 (cache, rate limits, sessions fallback).
Tooling
Docker Compose for one-command self-hosting; Celery for optional background jobs.
Language & quality
Type-safe API layer, Pydantic validation, structured config validation on boot.
API service
REST API with JWT auth, workspace-scoped routes, SSE streaming for chat.
AI Gateway
Central model-routing layer — single client, multiple providers.
MCP runtime
Serves tools over SSE JSON-RPC with permission checks and audit.
Agent runtime
Executes agents with recorded state, steps and tool outcomes.
Ingestion workers
Parse, chunk and embed documents asynchronously.
Gateway (future)
Standalone model-routing service for all connectors.
PostgreSQL 16
Users, workspaces, knowledge, conversations, agents, executions, usage — the system of record.
Redis 7
Cache, rate-limit counters, background queues.
Vector store
Embeddings live with your data today; pgvector-native storage is planned.
Object storage
Uploaded documents and generated artifacts in your storage.
Security & Privacy
Security is designed in, not bolted on
Local-first means your data, models and audit records can all stay inside your estate.
Identity
JWT access (15 min) + rotating refresh (7 d, HttpOnly cookie, path-scoped), Argon2id password hashing, token-bucket rate limiting per client.
Tenant isolation
Every entity is workspace-scoped; the X-Workspace-Id header selects the tenant for knowledge, agents, tools and models.
Authorization
Role model with rank ordering defined; route-level enforcement is the next milestone (today: ownership + per-object guards).
Tool safety
MCP tools enable/disable per workspace; no shell, OS or arbitrary code execution; high-risk actions audited.
Audit trail
Executions, tool calls, usage and system events recorded — a trail of what the AI did, when, and why.
Encryption & secrets
TLS at the proxy; env/config validation enforces rotated secrets; secrets never placed in model prompts.
Honest framing. Local-first reduces external dependency and improves data control, but it is not a guarantee of absolute privacy or security. Actual protection depends on how you deploy, configure and operate the platform.
Pricing
Platform pricing that scales with AI engineering
We price AI Forge as an AI engineering & agentic platform — not a chatbot. Local LLM inference is included when you run with your own infrastructure; third-party model costs are billed separately by the respective provider. Prices are indicative USD list prices.
Monthly billing · convert to annual for 2 months free on paid plans.
For developers, students, experimentation and POCs.
- 1 user · 1 project
- Local LLM / Ollama + Knowledge Base
- Basic RAG · PDF / DOCX / TXT / MD upload
- Basic AI Chat · basic embeddings
- 3 AI agents · 3 MCP tools
- Basic conversation memory
- Self-hosted Docker deployment · Community support
- MageTech AI Forge branding
For individual developers and AI builders.
- Everything in Community
- 5 projects · 1–3 users
- Unlimited knowledge bases
- 25 agents · 25 MCP tools
- Advanced RAG · agent memory
- LangGraph workflows · agent execution history
- Agent testing · API access · AI Gateway
- Document source citations · Basic AgentOps
- Custom prompts · Email support
For teams building internal AI applications together.
- Everything in Developer
- 10 users · 20 projects · unlimited agents & MCP tools
- Advanced AgentOps · team collaboration
- Role-based access (Admin / Manager / Developer / Viewer)
- Shared knowledge bases · advanced memory
- Agent execution monitoring · audit logs
- MCP permissions / allowlists · API + webhooks
- Multiple AI providers · advanced RAG
- Priority support
For companies building internal AI applications and business agents at scale.
- Everything in Team
- 25 users · unlimited projects, KBs, agents, MCP tools
- Advanced AgentOps · advanced RBAC · audit trail
- SSO · advanced security controls
- Private AI Gateway · multiple LLM providers
- Custom MCP & business-system integrations
- Usage analytics · advanced monitoring
- Custom retention policies · deployment assistance
- Priority support
For larger organizations requiring private AI infrastructure and full control.
- Unlimited users & projects
- Private / on-premise / Kubernetes deployment
- Dedicated AI infrastructure · private local LLM environment
- Custom MCP servers & custom integrations
- SSO / SAML · advanced RBAC · audit logs
- Security reviews · data-retention policies
- Dedicated support · SLA · architecture consultation
- Custom onboarding · custom contract / invoicing
AI model usage. Local LLM inference is included when you run MageTech AI Forge with your own infrastructure (Ollama). Third-party model/API costs are billed separately by the respective provider — we do not bundle or inflate cloud inference into platform pricing.
| Feature | Community | Developer | Team | Business | Enterprise |
|---|---|---|---|---|---|
| Monthly | $0 | $29 | $99 ⭐ | $249 | Custom |
| Users | 1 | 3 | 10 | 25 | Unlimited |
| Projects | 1 | 5 | 20 | Unlimited | Unlimited |
| Knowledge bases | 1 | Unlimited | Unlimited | Unlimited | Unlimited |
| Agents | 3 | 25 | Unlimited | Unlimited | Unlimited |
| MCP tools | 3 | 25 | Unlimited | Unlimited | Unlimited |
| RAG | Basic | Advanced | Advanced | Advanced | Custom |
| AgentOps | Basic | Basic | Advanced | Advanced | Enterprise |
| Memory | Basic | Yes | Yes | Yes | Yes |
| API access | — | Yes | Yes | Yes | Yes |
| RBAC | — | Basic | Yes | Advanced | Enterprise |
| SSO | — | — | — | Yes | Yes |
| Audit logs | — | — | Yes | Yes | Yes |
| Private deployment | Yes | Yes | Yes | Yes | Yes |
| On-premise | — | — | — | — | Yes |
| Support | Community | Priority | Priority | Dedicated |
A clean progression: Free → Developer → Team → Business → Enterprise. Free for POCs, Team is the sweet spot for most teams, Business adds SSO and scale, Enterprise is fully private and on-premise. We keep the public tier structure simple and add self-hosted licensing separately.
Self-hosting
Self-hosted software licensing
Because self-hosting is core to AI Forge, we separate the software license from your infrastructure. You bring the hardware (GPU/CPU, PostgreSQL, Redis, Ollama or local models, storage, networking); we provide the software and support.
| Edition | Price | What you get |
|---|---|---|
| Community | Free | Self-hosted Docker, Community support |
| Developer License | $299 / year | Email support, commercial use |
| Team License | $999 / year | Priority support, MCP allowlists, audit logs |
| Business License | $2,499 / year | SSO, advanced RBAC, deployment assistance |
| Enterprise | Custom | On-prem, SLA, dedicated support, security reviews |
MageTech provides the AI Forge software and support under the license; you own and operate your infrastructure.
India
India pricing (INR)
Localized pricing for the Indian market. USD remains the primary international price; INR is an approximate regional equivalent.
| Plan | USD | Approx. INR |
|---|---|---|
| Community | $0 | ₹0 |
| Developer | $29 / mo | ₹2,499 / mo |
| Team | $99 / mo | ₹8,499 / mo |
| Business | $249 / mo | ₹20,999 / mo |
| Enterprise | Custom | Custom |
India pricing shown is indicative and subject to confirmation. Annual billing applies where noted.
Why separate pricing? It keeps economics safe: local inference is included on your own infrastructure, cloud inference is metered by the provider, and the software license is priced like the AI infrastructure market — open core at the base, enterprise capabilities and support sold separately.
Roadmap
Built in phases, honest at each step
The roadmap separates the product into clear milestones so you always know what is live, what is being built, and what comes next.
Knowledge / RAG foundation Live
Knowledge bases, ingestion, chunking, embeddings, semantic search with citations, multi-turn chat and workspace isolation.
MCP / Agents Live
Agent runtime, MCP servers & tools over SSE, tool permissions, execution history, AI Gateway with auto / ollama / mock modes.
AgentOps · Memory · Security In progress
Advanced AgentOps, agent memory, RBAC enforcement, refresh-blacklist hardening, webhooks, SSO/OAuth2 and expanded audit surface.
Business integrations Future
Native pgvector, LangGraph workflows, CRM / ERP / helpdesk connectors, managed providers (OpenAI / Gemini / Anthropic) and deeper MCP / ServiceFlow integration.
Not a promise. Roadmap items are direction, not commitments — they can change. Readiness badges always reflect current facts in the repository, not aspiration.
FAQ
Frequently asked questions
Straight answers to the questions we hear most often.
Does everything run on my own hardware?
Yes. The full stack — API, knowledge, RAG, agents, MCP and the AI Gateway — can run self-hosted via Docker Compose with Ollama for local LLM inference. Nothing is forced through a cloud service.
Which models does it use?
By default `qwen2.5:7b` for generation and `nomic-embed-text` (384-dim) for embeddings, served through Ollama. The AI Gateway routes any configured model, so you can point it at additional local models today and managed providers in the future.
How are answers grounded and cited?
Documents are parsed, chunked (~800 chars, 160 overlap), embedded and stored. On a query, semantic search retrieves top chunks (cosine similarity), the context is passed to the LLM, and the answer cites its sources so you can verify.
Can agents connect to my business systems?
Through MCP. Tools are exposed over SSE JSON-RPC, enabled per workspace, permission-checked and audited. There is no shell / OS / arbitrary code execution via AI tools — a deliberate safety boundary. Business-system connectors (e.g. ServiceFlow) are roadmap items.
Do you support OpenAI, Gemini or Anthropic?
Not yet — they are planned connector providers on the AI Gateway. Local models remain the default, which is exactly why the pricing keeps local inference included and cloud inference billed by the provider.
Is my data private?
Local-first design means your knowledge, models and audit records can stay inside your estate. That reduces external dependency and improves data control — but it is not an absolute guarantee of privacy or security; protection depends on how you deploy and operate the platform.
What does the pricing include?
Platform access, self-hosted deployment, agents / MCP / RAG entitlements per tier, and support. Local LLM inference on your own infrastructure is included; third-party model API costs are billed separately by the provider. Prices shown are indicative.
Can I deploy on-premise or in Kubernetes?
Community through Business self-host via Docker Compose on your own servers. On-premise and Kubernetes deployment, dedicated AI infrastructure and a private local LLM environment are Enterprise offerings.
Get started
Ship your first AI application for free
Start on Community, grow into Team, and upgrade to Business or Enterprise when SSO, advanced RBAC and scale become priorities.