AICortex

[AICortex — your self-hosted LLM brain, on your own NAS]

[License: Apache 2.0] [Latest release] [Build] [Protocol: MCP] [Self-hosted] [Buy me a coffee]

A private, self-hosted brain for your LLM — on your own NAS.

In 30 seconds: AICortex is one container on your own hardware that every AI you use connects to.

• It remembers who you are, how you work and what you're building — and hands that to Claude, ChatGPT or a local model at the start of every chat, on desktop and phone.
• It reaches your stuff — services, devices, mail, calendars, files — with the secrets in an encrypted vault the model never sees.
• It keeps your schedule: tell your AI "tomorrow at 8: summarise yesterday's notes". AICortex stores the job and at 8 wakes your Claude (or asks ChatGPT, an API key or a local Ollama) to do it.

[Scheduled jobs in the AICortex WebUI: a morning briefing on Claude, a weekly review on an OpenRouter free model, a one-off backup check and an evening job on a local Ollama, each with its next and last run]

Try it in 5 minutes · Run it on your NAS · More screenshots · Connect your AI

AICortex is a self-hosted MCP server that turns your NAS into a personal LLM connector: a persistent "brain" your assistant loads at the start of every session. Any MCP-capable LLM — Claude, ChatGPT, or a fully local model — gains a durable identity and real reach into your own services and devices, while running inside your network. Non-MCP tools (n8n, LangChain, your own scripts) reach the same brain through a native REST API with scoped per-user keys — and humans get a built-in admin WebUI at /ui for the vault, skills, users and audit log.

The model stays in its provider's cloud (or runs locally). Your memory, skills and secrets stay on your NAS. The assistant talks to AICortex over an HTTPS connector; the server uses your local credentials internally and never hands them to the model. New capabilities are added as data — a skill, a service config, a secret — with no redeploy.

Highlights

🧠 A brain that travels with you

• One short cortex_start call — who you are, how you work, and the current project's notes, open tasks and last session (a few hundred characters; cortex_catalog has the full list)
• Self-learning, typed memory with a review queue, so it grows without polluting itself
• A skill router — search your skills, load only what's needed, author new ones at runtime
• Playbooks — multi-stage work where a person approves each stage's artifact (e.g. Anthropic's AI-native SDLC: intent.md → spec.md → plan.md …), bundled as a template — see docs/playbooks.md
• The same brain across models, desktop and mobile — one account, one state

🔌 Real reach — every integration is just data

• Email (SMTP send + IMAP read), calendars (CalDAV), HTTP services, MQTT & FTP/FTPS devices, WebDAV cloud, SSH/SFTP
• IPP printing & eSCL scanning (straight into Paperless-ngx) on LAN multifunction devices
• Event-driven: inbound webhooks (POST /hooks/<name>, secret/HMAC-verified) turn external events into inbox items; outbound webhooks notify
• An MCP gateway to use other MCP servers' tools — register any integration with one call, no code, no redeploy

🌐 Beyond MCP — drive it over plain HTTP

• A native REST API: every tool as POST /api/v1/tools/<name> with an auto-generated OpenAPI 3.1 spec — plug AICortex into n8n, LangChain, an OpenAI-compatible client or a shell script, no MCP client required
• Per-user API keys: scoped (default-deny), hashed at rest, rate-limited, optional expiry — and a key runs the same per-user areas as an OIDC session, never admin by default

🖥 Built-in admin WebUI — no terminal needed

• Every signed-in person manages their own notes, scheduled jobs and playbook runs (and approves playbook stages) at /ui; admins also get the vault (secret names only — values are write-only), skills, users & roles, services & devices editing and the audit log — language-switchable DE/EN
• Same SSO as the connector (OIDC + PKCE against your IdP, e.g. Pocket ID), CSRF-guarded, fully self-contained (strict CSP, no CDNs, no build step) — served by the same container, no extra service
• Every page takes the same policy path as the MCP tools, so the browser never gets more than the person's AI
• See it in screenshots

👥 Team, autonomy & continuity

• A presence-aware multi-agent board: live presence, capability-routed task pull, context-preserving handoff
• Cross-LLM session handoff — resume exactly where another model or device left off
• Cron-as-data scheduling with per-user act-as: a scheduled job runs confined to its owner's area, via a short-lived per-job token — never a standing god-token
• Runners carry the jobs out — AICortex wakes your Claude (a Claude Code cloud task on your plan) with the due job, or asks your ChatGPT plan, an API key or a local Ollama; each person sets up their own in the WebUI and gives each job its runner, no extra container (how)

🔒 Secure & multi-tenant

• Your own OIDC login and roles (admin / user / viewer) — one switch, AUTH_ENFORCE
• Per-user isolation and capability areas: private memory + vault per person, plus default-deny service/skill allow-lists an admin assigns — so not everyone may use every tool
• An encrypted secret vault, an SSRF egress guard, and capability checks that fail closed — end-to-end on your hardware, including a local model (Ollama)

How it works

LLM app — any MCP client (desktop / mobile)  ·  one or many agents        │  custom connector / MCP server (HTTPS, from the model's cloud)        ▼Reverse proxy (Zoraxy / Caddy / nginx / Traefik …)        │        ▼AICortex  (this container, on your NAS)        │  uses local files & secrets        ▼Memory · Skills · HTTP services · MQTT & FTP devices · WebDAV cloud · Workspace files   · SSH/SFTP · SMTP + IMAP email · IPP printing · eSCL scanning · MCP gateway · Inbox/Tasks   · Sessions · Cron · Secret vault   (every outbound call passes the SSRF egress guard)

How AICortex differs

The "LLM brain" space has some great projects — AICortex deliberately sits in a different spot:

• GBrain is a memory engine — markdown-first, Postgres/PGLite-backed, with a self-wiring knowledge graph and hybrid/vector search. Deep retrieval, one concern.
• CortexPrism is a full agent operating system — its own runtime, ~30 LLM providers, a plugin marketplace, sandboxed code execution and multi-user teams. Batteries included, heavyweight.
• AICortex is a single, lightweight MCP connector: one container that gives any client a persistent brain — memory and skills, devices, sessions, secrets and scheduling — where new integrations are added as data, no redeploy. And unlike most brain projects it isn't MCP-only: the same tools are exposed over a native REST API with an auto-generated OpenAPI spec, so non-MCP clients (n8n, LangChain, a script) plug in too. No database, no graph, no marketplace; you bring your own client.

Rule of thumb: want a dedicated memory/graph engine → GBrain; want a batteries-included agent OS → CortexPrism; want a minimal self-hosted connector that plugs your existing LLM app into your own stuff → AICortex.

Try it in 5 minutes

Just want to look around? Run it on your own computer — no NAS, domain or login needed, only Docker:

echo "STORAGE_ENCRYPTION_KEY=$(openssl rand -base64 32 | tr '+/' '-_')" > aicortex.envdocker run -d --name aicortex -p 127.0.0.1:8787:8787 --env-file aicortex.env \  -e ALLOW_INSECURE=1 -v aicortex-data:/data ghcr.io/ikarusmk/aicortex:latest

1. Open http://localhost:8787/ui — the WebUI, already signed in as admin.
2. Connect an AI that runs on the same computer. Claude Code: claude mcp add --transport http aicortex http://localhost:8787/mcp, then ask it to call cortex_start. Other local clients: docs/clients.md. The Claude and ChatGPT apps connect from their provider's cloud, so they need the real setup.
3. Tell it something about you, or "tomorrow at 8: summarise yesterday's notes" — it shows up under Notes and Scheduled jobs. A job runs once you add a runner with an API key (OpenRouter's free models work); waking Claude needs your own login, so that, too, is the real setup.

⚠️ No login in this mode. ALLOW_INSECURE=1 switches authentication off: whoever reaches the port is admin. Keep the port on 127.0.0.1 as above, keep real secrets out, and never run it like this on a server. For everyday use, follow the Quick start with your own login.

Done looking? docker rm -f aicortex && docker volume rm aicortex-data && rm aicortex.env removes it with all its data.

Requirements

• A NAS or server running Docker (Compose v2).
• A reverse proxy serving the container over public HTTPS (cloud-hosted LLM clients connect from their provider's cloud, so the endpoint must be reachable from the internet; a purely local client can reach it on the LAN).
• A domain/subdomain pointing at your proxy.
• An MCP-capable client that supports custom connectors / MCP servers (Claude, ChatGPT, or any MCP client).

Quick start

🤖 Installing with an AI coding agent? Point it at INSTALL_FOR_AGENTS.md — it walks the agent through the whole install and verification, handing the human-only steps (DNS, proxy, OIDC) back to you.

git clone [email protected]:IkarusMK/AICortex.gitcd AICortexcp .env.example .env        # adjust PUID / PGID / HOST_PORT / TZdocker compose up -d --build

The MCP endpoint is served at http://<host>:8787/mcp.

Prebuilt image (no local build): a multi-arch image is published to GHCR by CI. In docker-compose.yml, comment out build: ., uncomment image: ghcr.io/ikarusmk/aicortex:latest, then docker compose pull && docker compose up -d.

Updating from 1.x: pull the new image and restart with the same data/ folder and .env — there is no migration step. Connected apps stay signed in; memory, skills, vault, services, tasks, sessions and API keys carry over. What changes is listed in CHANGELOG.md.

Expose it & add the connector:

1. Point a subdomain (e.g. agent.example.com) at your reverse proxy.
2. Proxy that host to http://<nas-ip>:8787 over HTTPS — the upstream is plain HTTP, so do not enable "TLS to upstream". If your proxy geo-blocks, allow your LLM provider's egress region. The published port binds to 127.0.0.1 by default (only this host's proxy reaches it, not the LAN); a host-network proxy uses http://127.0.0.1:8787, a bridge-network proxy needs BIND_ADDR=0.0.0.0 (or a shared Docker network).
3. In your MCP client: add a custom connector / MCP server → URL https://agent.example.com/mcp. Agents and CLIs that can only send a header (Claude Code, Codex, Hermes, OpenClaw, LM Studio, …) use a per-agent key instead — setup for every client: docs/clients.md.
4. Test: ask the assistant to call cortex_start. The briefing starts with AICortex v<version>; on a fresh install it is short because the memory is still empty. (ping works too, but it is not among the eleven tools clients see up front.)

Authentication & authorization

AICortex separates authentication (who may connect) from authorization (what they may do). Set up the login once, then pick your mode with a single line in .env.

1. Authenticate with your own OIDC provider

Exposing the connector publicly requires OAuth. AICortex uses your own identity provider as the login backend — Pocket ID, Authentik, Keycloak, Auth0, anything with standard OIDC discovery. FastMCP's OIDC proxy handles the MCP-side OAuth 2.1 flow (Dynamic Client Registration + PKCE); your provider does the actual login.

Register the OAuth client in your provider with redirect URI <BASE_URL>/auth/callback, then (re-)add the connector in your client to be sent through the login.

🟣 Using Pocket ID? Follow the click-by-click Pocket ID setup guide — creating the OIDC client, the exact .env, and optional groups → roles.

ℹ️ Don't put reverse-proxy SSO / forward-auth in front of /mcp — an MCP connector is a machine client and can't follow an interactive login redirect. Authentication belongs at the MCP layer, which is exactly what this does. When the OIDC variables are unset, /mcp accepts bearer tokens only (per-agent keys and RUNNER_TOKEN); ALLOW_INSECURE=1 turns all auth off for local testing.

2. Pick your mode — one line in .env

🏠 Homelab — keep it simple

AUTH_ENFORCE=0

Every authenticated caller gets every tool — one shared brain, zero friction. Authentication (OIDC) still guards the front door; you've just turned off the internal role gate.

🏢 Enterprise — several people, one brain

AUTH_ENFORCE=1                          # default — the single switch (see below)OIDC_SCOPE=openid profile email groups  # request the groups claimAUTH_ROLE_CLAIM=groups                  # map Pocket ID groups → roles

AUTH_ENFORCE=1 is the one switch — "enforce means enforce". It turns on roles (admin / user / viewer), per-user isolation, and per-user capability areas together. Roles are driven by your IdP groups (mapped in data/auth/policy.json, or managed live with the tenancy_* admin tools). Each non-admin is confined to their own memory scope and private vault namespace — two people never read or overwrite each other's data — and reaches only the services & skills an admin assigns (default-deny); an admin provisions per-user secrets (secret_set owner=…) and areas (tenancy_set services=… skills=…). Admins keep full access.

Both modes run the same image. Private data (memory, vault) fails open so a glitch never strands the operator; shared capabilities (services, skills) fail closed under enforce. Full guides: docs/authorization.md · docs/per-user-areas.md.

Capabilities

Every register has a matching delete (skill_delete, service_delete, mqtt_delete, memory_delete, secret_delete, …), so anything you register you can also remove via the connector.

Memory, skills & the skill router

This is the heart of the project — making the assistant itself portable, not just the chat.

• Memory lives as plain files under data/memory; the memory_* tools let the LLM recall and update what it knows about you, the same on every device. The catalog groups it by tier — 🧭 Core (identity & preferences) → 📂 Projects → 🛠 Working style → 🔗 References, derived from each memory's type, with short-term/current state in the sessions layer — so a fresh LLM sees who you are before the ephemera.
• Skills live as folders under data/skills (<skill>/SKILL.md + resources). The router (skill_search / skill_load / skill_resource) finds the right skill and pulls in only what it needs (progressive disclosure).
• Categories keep it cheap — and are mandatory. skill_write refuses an uncategorized skill and snaps near-duplicate spellings onto the existing category, so a 300-skill library stays a dozen lines in cortex_catalog, not a token dump. Starter skills live in examples/skills/ and ship with the image — Skills → Add starter skills in the WebUI adds them.
• cortex_start first. The server instructions ask every model to call it at the start of a conversation (with the project name). The briefing names every general working rule (a feedback memory without a project) — tag a rule pinned to put it first with its text — while rules that belong to a project only come along for that project. For clients that don't pass the instructions on, add a short block to your client's project/system prompt — see docs/client-project-instructions.md.

Auto-memory — a brain that learns by itself

Most assistants forget the moment a chat ends. AICortex closes that loop — and stays tidy while doing it.

• Typed & tiered. Every memory is user, feedback, project or reference; memory_write refuses an untyped memory, and cortex_catalog groups them into tiers (Core → Projects → Working style → References) so the catalog reads long-term → ephemeral — the brain stays sorted by intent.
• Learns in-session, at zero extra cost. The model already talking to you distills the durable facts and writes them back — no second model, no background loop.
• Dedup-first. A write flags overlapping entries so related facts merge into one file instead of multiplying.
• A review queue. Anything captured automatically (or staged with memory_note) lands as a candidate, not live memory — you memory_promote the keepers and memory_reject the rest. Set LEARN_AUTOCAPTURE=0 to turn auto-capture off.

Tools & integrations (as data)

New integrations don't need new code. A service is a small config you register at runtime with service_add (stored under data/services); call_service then reaches it — only registered services are allowed, and the auth token is injected server-side from the vault (token_env), never stored in service data or returned to the model. Every service takes a category (like skills — service_add refuses without one), so the catalog stays grouped and findable. Pair a service with a skill_write skill that explains how to use it, and a new capability is live without a redeploy.

Multi-agent — one brain, many agents

Run your assistants as a team: Claude on the desktop, a local model on the NAS, your phone — sharing one brain and one task board.

• Live presence. agent_register doubles as a heartbeat; agent_list shows who's online / idle / away, and cortex_catalog surfaces the team.
• Pull work, don't hunt for it. task_next(owner) recommends the best open task (assigned-to-you → capability match → unassigned); task_claim takes it.
• Hand off with full context. task_handoff(id, to) reassigns a task, notifies the recipient, and attaches the work session — they session_load and pick up exactly where the other stopped.
• Namespaced memory. Shared knowledge in shared, private notes in agents/<name>.

Sub-agent spawning stays client-side; the connector is the shared coordination layer they meet on. See docs/ARCHITECTURE.md.

Autonomy & scheduling

New here? → Scheduled jobs and runners, step by step.

Schedules live on the NAS as data — cron_add(name, schedule, prompt[, runner, once]) from any device, or Scheduled jobs in the WebUI (presets, a preview of the next runs, on/off, the job's runner). Runners carry them out — set up in the WebUI, tab Runners, by every person for their own jobs; no extra container, nothing to install:

1. Once a minute, on the minute, AICortex checks for due jobs — without any model, so an idle minute costs nothing. A minute it missed (slow run, restart) is caught up for five minutes.
2. Only when a job is due does it wake or ask the job's runner. A job that names none goes to its person's own default runner (their first one, or the one they chose), else to the standard runner (one of an admin's). A one-off job (once) switches itself off after its run.
3. The job runs through this connector and reports via your configured channel or the connector inbox; cron_mark_run records the result.

Kinds of runner:

A key typed in is stored encrypted as the runner's own vault entry (RUNNER_KEY_<id>, an internal name no service can use) and never shown again; picking an existing entry (e.g. OPENROUTER_API_KEY) needs no second copy — a person may pick only their own entries, an admin shared ones too. Where AICortex asks the model itself, each job is a conversation of its own: AICortex confines it to the job owner's area (act-as) before the model sees it, ends that afterwards and records the run — the model never sees a token or the runner plumbing; its tool calls go through /mcp, so roles and audit apply as for any caller. A job gets five minutes; runners work side by side, so a slow one holds up nobody else. Test on a runner asks its model one tiny question, or wakes the cloud task with a harmless check and shows the link to the Claude session. Limits: 10 runners per person (admins: no limit), keys up to 1,024 characters, job prompts up to 20,000.

A Claude runner acts as you. The cloud task runs with your full rights in AICortex and all your claude.ai connectors, so a Claude runner only carries out jobs you wrote yourself — their prompt, name and recipient — and never ones written with an agent key. Anthropic marks the wake-up text as untrusted data that can't grant approval — like every runner, a job does only what is read-only or explicitly allowed in its prompt. Limits: 30 wake-ups an hour per task (AICortex keeps two minutes between them), no automatic retry (a retried wake-up would start a second session). Why not claude setup-token? Anthropic refuses a subscription token outside Claude Code (HTTP 429 "Error"), so 2.3.0 dropped that way; a 2.2 setup using it becomes a switched-off Claude runner.

The ChatGPT plan (Plus or Pro) uses OpenAI's Sign in with ChatGPT for open-source apps: choose ChatGPT plan, click Sign in with ChatGPT, open the sign-in and allow AICortex to use the plan. OpenAI then sends the browser to http://127.0.0.1:1455/… — the computer you sit at, not the NAS — so the page shows an error; paste that whole address into the field below and finish. AICortex checks the ID token, keeps the tokens encrypted in the vault, refreshes them itself (the sign-in stays valid as long as AICortex renews it within 30 days, which it does on its own) and runs jobs through OpenAI's Responses API. Leave the model empty for your account's first model, or pick one of the suggestions. On Plus, all apps share the plan's five-hour limit; Sign out revokes the sign-in at OpenAI.

An own runner outside AICortex that calls cron_due with the static RUNNER_TOKEN from the connector's .env keeps working: it gets the jobs assigned to external, and the unassigned ones while no standard runner is on. A job is handed out once per scheduled minute, so two runners never run it twice.

What the runner may do. In homelab mode (AUTH_ENFORCE=0) it has full access. In enterprise mode (AUTH_ENFORCE=1) it is a least-privilege user: it can't register integrations or set secrets, and reaches only the services/skills an admin assigned. For a runner that should act broadly, set RUNNER_ROLE=admin; for scheduled work on a person's behalf, give the job an owner so it runs act-as in that user's area (short-lived per-job token, no standing authority). See docs/authorization.md.

Plan or API key? A Claude cloud task and the ChatGPT sign-in are the providers' own ways for this and count against your plan's limits; an API key bills per use. Because a runner is only woken when a job is due, the usage stays small either way.

Run it with a local model (Ollama)

AICortex is model-agnostic — including fully local models. LibreChat is a self-hosted chat UI with first-class MCP support, Ollama runs the model locally, AICortex is the brain/tools — nothing leaves your network. LibreChat connects to your existing AICortex over MCP with per-person OAuth (the same OAuth 2.1 + dynamic-registration flow Claude.ai uses) — no connector changes needed.

Browser ─▶ LibreChat (chat UI + MCP client)  ──MCP, per-person OAuth──▶  AICortex  ──▶  memory · skills · tools                 ▼              Ollama (local model)

In short: make Ollama reachable on your LAN (OLLAMA_HOST=0.0.0.0), bring up the LibreChat stack (its own compose — never merged into AICortex's), add AICortex as an MCP server (type: streamable-http, requiresOAuth: true), authenticate once through your IdP, and drive it through a curated Agent so a mid-size local model isn't overwhelmed by 100+ tools.

Pick a tool-capable model (recent Qwen/Llama-class; small <7B models struggle with multi-step tools). Full step-by-step guide — compose, .env, the OAuth connect, troubleshooting, and optional cloud models (OpenRouter, Ollama Cloud): docs/local-llm-librechat.md.

FAQ

Why a separate "brain" when Claude and ChatGPT already have memory? Because processing your data isn't the same as owning it. Built-in memory lives inside one vendor, one app, one model. AICortex is yours: model- and client-agnostic (the same brain on Claude, ChatGPT, a local Ollama model, desktop or phone — no lock-in), it reaches your own devices and files, and it keeps secrets in a vault on your NAS. Its action tools reduce what you paste into a cloud chat rather than add exposure — the sensitive value stays server-side and is referenced by name.

Why keyword search instead of semantic / vector search? Deliberate. Memory is a curated, typed catalog — the handful of facts you (or the assistant) chose to keep — not an embedding pile you hope to retrieve from. A typed, human-readable catalog stays debuggable, portable and free of a vector database. (If the skill library ever grows large enough that paraphrase-matching hurts, opt-in local embeddings for skill_search only is the escape hatch — not a memory-wide graph.)

Why no built-in code sandbox / executor? A code executor would be the single biggest attack surface and it breaks the no code, no redeploy principle. The sandbox pattern already exists as data: ssh_add a disposable, LAN-isolated jail host and ssh_run there. It's a documented pattern, not something baked into the connector.

Why no built-in Telegram / Slack / ntfy / … integrations? Those are just HTTP APIs — so they're service_add + a skill, not new code. That's the whole point: a new capability is data + a skill, never a redeploy.

Do I need OIDC and a reverse proxy? For a public/cloud LLM client, yes — the endpoint must be reachable over HTTPS and authenticated (see Authentication & authorization). Running a purely local client? You can keep it on the LAN. Homelab mode keeps auth simple; enterprise mode adds roles + per-user isolation with one switch — see docs/authorization.md.

Security

• Auth fails closed. Without OIDC, /mcp accepts only bearer tokens (RUNNER_TOKEN or a per-agent key); ALLOW_INSECURE=1 runs it open for local development only. Enable OIDC before exposing the proxy for Claude or ChatGPT.
• DNS rebinding blocked. Browser requests (and the WebUI) must come from an IP address, localhost, a name without a dot, a private-use name or the BASE_URL host; native clients are unaffected. MCP_ALLOWED_HOSTS makes the host check strict.
• SSRF guard. service_add / mqtt_add / ftp_add are model-callable, so the registered-target list isn't a trust boundary by itself. Every outbound host is resolved and private / loopback / link-local / cloud-metadata addresses are blocked unless they fall inside INTERNAL_ALLOW_CIDRS (operator-only). Set it to the LAN/VPN ranges you actually use.
• Encrypted vault, enforced. Secrets go in the vault via secret_set (encrypted at rest, referenced by name, never returned). secret_set refuses plaintext unless STORAGE_ENCRYPTION_KEY is set. .env is only for bootstrap config.
• Inbound webhooks authenticate themselves. POST /hooks/<name> is served alongside /mcp/ but not behind the MCP OAuth (external senders can't do OAuth) — instead each hook validates a shared secret token (the X-Webhook-Token header only — never a URL query, so it can't leak into logs) and/or an HMAC signature (constant-time). Unknown hooks and bad/unsigned requests both return a uniform 401 (no hook-name enumeration), the body size is capped, and it only ever deposits into the inbox — it never reaches the tool surface. Expose only /hooks/* past your reverse proxy's auth, never /mcp. Don't want any public receiver? Leave webhooks unregistered (opt-in) or keep the /hooks path LAN-only.
• TLS verified by default for FTP / MQTT / WebDAV / scanning; self-signed LAN devices opt out per-endpoint via the admin-only registration tools (tls_insecure / ca_bundle).
• .env and data/ contents are git-ignored — never commit secrets.

AICortex has been through external security review (fail-closed vault, SSH host-key pinning, connect-time DNS-rebinding protection, resource limits, the authorization layer and per-user isolation). Full status & reporting: SECURITY.md.

🔐 How your secrets stay secret — encrypted at rest, never returned to the model, never in chat/repo, and how to set them safely (secret_set or .env): docs/secrets.md.

REST API (for non-MCP clients)

Not every client speaks MCP. The native REST layer exposes the same tools over plain HTTP, so n8n, LangChain, an OpenAI-compatible client or a shell script can drive AICortex directly — through the same authorization and per-user areas as an OIDC session.

Three routes, served next to /mcp but authenticated by a per-user API key instead of OAuth:

An admin mints keys (they run through the same policy, never admin by default):

apikey_create(identity="alice", scopes="memory, fs_read", ttl_days=90)# → ak_ab12cd34ef56_XXXXXXXX…   (shown ONCE — store it now)apikey_list        # keyid prefixes, scopes, expiry — never the secretapikey_revoke("ab12cd34ef56")

Call it with the key as a bearer token:

curl -X POST https://agent.example.com/api/v1/tools/memory_search \  -H "Authorization: Bearer ak_ab12cd34ef56_XXXXXXXX…" \  -H "Content-Type: application/json" \  -d '{"query": "trading rules"}'

Security, by design: keys are hashed at rest (constant-time compare) and default-deny — a key reaches only the tools in its scopes allow-list, further narrowed by the identity's role and device areas. Secret-, key- and tenancy-management tools are never reachable via a key. Each key is rate-limited; pass ?stream=1 (or Accept: text/event-stream) for SSE on long-running tools. Full detail: docs/rest-api.md.

Proxy: expose /api/* (and /hooks/*) past the reverse proxy without OIDC — the API authenticates itself by key. Never expose /mcp that way.

Admin WebUI

Not everyone wants to manage secrets from a terminal or chat. /ui is a built-in, self-contained admin panel (no extra container, no build step, no CDNs — strict CSP), language-switchable DE/EN:

[Playbooks: a run of the AI-native SDLC template — the plan stage approved, the design stage's spec.md waiting for approval]

_{Screenshots use made-up demo data.}

Login rides on your existing IdP (authorization code + PKCE): register one extra redirect URI in the OIDC client — <BASE_URL>/ui/callback — and open <BASE_URL>/ui. Everyone works in their own area; the management pages are for the admin role. Sessions are signed HttpOnly cookies, mutations are CSRF-guarded, every action lands in the audit log (names only, never values). Each page runs the same policy path as the MCP tools, so the browser gets exactly the rights of the person's AI. Without OIDC (local testing) the UI is open on the localhost-only bind — the same trust model as the MCP endpoint itself. There is no separate username/password login: remote access always goes through your IdP. Opt-out: UI_ENABLED=0.

First steps after the install — sign in, set up people, create keys and connect Claude, ChatGPT, Claude Code and other agents: docs/clients.md → Step by step.

Configuration

All config lives in .env (copy from .env.example):

OIDC, authorization and hardening variables are documented inline in .env.example and the Authentication & authorization section.

Deploying on a VPS or over a VPN

The connector is just an HTTPS MCP endpoint, so it runs anywhere Docker does. The SSRF guard keeps that safe:

• VPS, public APIs only: same reverse proxy + OIDC, set BASE_URL to its public URL, leave INTERNAL_ALLOW_CIDRS empty.
• Reach home devices from a VPS: link the VPS and your LAN with a VPN (WireGuard, Tailscale, …) and add the VPN/remote subnet to INTERNAL_ALLOW_CIDRS (e.g. 100.64.0.0/10 for Tailscale).
• Multi-site: scope FORWARDED_ALLOW_IPS to the proxy, list only trusted subnets in INTERNAL_ALLOW_CIDRS, and use a dedicated OIDC client.

Project structure

AICortex/├── app/                # Server code (FastMCP) — one module per tool group│   ├── server.py       #   entrypoint — wires auth + registers tool modules│   ├── memory.py · skills.py · services.py · mqtt_tools.py · ftp_tools.py│   ├── webdav_tools.py · fs_tools.py · ssh_tools.py · mail_tools.py│   ├── print_tools.py · scan_tools.py · mcp_gateway.py · coordination.py│   ├── cron.py · sessions.py · secrets_store.py · bootstrap.py · guide.py│   ├── runners.py · runner_builtin.py · runner_cloud.py · chatgpt_signin.py   #   runners for scheduled jobs│   ├── authz.py · tenancy.py · pocketid_proxy.py · netguard.py · learn.py│   ├── webui.py            #   admin WebUI (/ui) — OIDC+PKCE browser login│   ├── webui_static/       #   its self-contained frontend (no build step)│   └── requirements.txt├── data/               # Persistent, human-readable state (git-ignored content)│   ├── memory/ · skills/ · services/ · mqtt/ · ftp/ · webdav/ · ssh/ · mail/│   ├── coordination/ · cron/ · sessions/ · printers/ · scanners/│   ├── vault/          #   encrypted secrets│   ├── auth/           #   OAuth client registrations + policy.json│   └── work/           #   workspace file hub (scans, downloads, fs_*)├── docs/               # Architecture, authorization, local-LLM & client guides├── Dockerfile · entrypoint.sh · docker-compose.yml · .env.example

Dependencies are baked into the image at build time — no install-on-start volume. data/, logs/ and secrets/ keep their structure via .gitkeep; their contents are git-ignored.

Troubleshooting

See what's wrong first

A handful of read-only commands surface almost everything. Replace the placeholders (<compose-dir>, <device-ip>) with your own values.

# 1) Container logs — startup banner + live errorsdocker logs --tail 50 aicortex
# 2) Verbose auth logging (then re-read the logs)#    docker-compose.yml → environment:   FASTMCP_LOG_LEVEL: "DEBUG"
# 3) Which version / feature is actually running in the container?docker exec aicortex python -c "import netguard; print('has tls_verify:', hasattr(netguard,'tls_verify'))"
# 4) Authorization decisions — who called what, allow/deny (per-person 'sub')tail -n 20 <compose-dir>/data/auth/audit.log        # set AUTH_AUDIT_ALL=1 to log allowed calls too
# 5) Can the container actually reach a device? (verify off = ignore self-signed)docker exec aicortex python -c "import httpx; print(httpx.get('https://<device-ip>/', verify=False, timeout=5).status_code)"

After pulling a new image, restart your LLM client session — don't just "refresh tools". MCP clients freeze the tool list at connect, so new tools/parameters only appear on a fresh session. If a new parameter still doesn't show after docker compose pull && up -d --force-recreate, confirm the container is really on the new image with command #3 above.

Common issues & fixes

Support ☕

AICortex is free and open source, and it will stay that way — no paid tier, no features held back.

If it keeps your assistant's memory going and you feel like saying thanks, a coffee is always gladly received:

[Buy me a coffee]

Not in the mood for coffee? These help just as much:

• ⭐ A star — so others looking for a self-hosted brain find the project.
• 🐛 An issue — when something is wrong, unclear or could be better.
• 💬 A word to a friend — who wishes their assistant would finally remember them.

License

Apache License 2.0 © 2026 IkarusMK