DocsGPT runs in four postures — Cloud, Managed, On-premises and Air-gapped — plus a free self-hosted community route. Every posture is available from one seat; the platform, licence (MIT) and features are the same in each. Some vendors gate on-premises behind thousands of seats; DocsGPT does not. A large Central European bank runs it fully air-gapped on its existing NVIDIA T4 GPUs.
MIT licence · v0.21.0 · every posture from one seat
Last updated: · Facts checked on
Cloud
Managed
On-premises
Air-gapped
Cloud: Arc53 operates a multi-tenant service and the LLM runs at a cloud provider. Managed: Arc53 operates a dedicated instance in your chosen region; the LLM runs at a cloud provider, on a local engine on the instance, or with your own key (BYOM). On-premises: you operate DocsGPT inside your network; the LLM runs locally, or (hybrid) prompts cross the boundary to a cloud provider you choose. Air-gapped: zero external network; LLM, embeddings, vector store and storage all run inside your perimeter.
24 h → 16 min
A large Central European commercial bank processes 10,000+ inbound documents a day; time-to-action fell from 24 hours to 16 minutes, with over 99.5 % classification accuracy, running fully air-gapped on the bank's existing NVIDIA T4 GPUs.
Source: EU bank, name under NDA
5,000+ documents · flat monthly fee
Dedicated managed instance for a specialist consultancy: 5,000+ documents, SSO, one flat monthly fee with no per-seat licence.
Source: No Deviation
18.3k GitHub stars
Open source under the MIT licence, with 226+ contributors. The same code runs in every posture.
Source: GitHub
Same platform, same MIT licence, same features. The differences are who operates it, where data lives, where the model runs and how it is priced.
Cloud and Managed are operated by Arc53 — Cloud on shared infrastructure in the EU region, Managed on a dedicated server in the Americas, Europe, Middle East, Africa or Asia. On-premises and Air-gapped are operated by the customer with Arc53 support, inside the customer's own network; Air-gapped adds a zero-egress rule and a local-only LLM. Cloud is priced per seat at $20/seat/month; Managed is a flat fee from $2,000 a month; On-premises and Air-gapped are from $10,000, one-time or recurring, with no per-seat licence. Every posture is available from one seat.
| Cloud | Managed | On-premises | Air-gapped | |
|---|---|---|---|---|
| Who operates it | Arc53 | Arc53 | Customer, with Arc53 support | Customer, with Arc53 support |
| Where data lives | Arc53 multi-tenant cloud, EU region | Dedicated VPS operated by Arc53 in the Americas, Europe, Middle East, Africa or Asia | Customer data centre or private cloud | Customer network, no external egress |
| Where the LLM runs | Cloud provider via Arc53 | Cloud provider via Arc53, a local engine on the instance, or your own provider key (BYOM) | Local engine (llama.cpp · Ollama · TGI · SGLang · vLLM · Aphrodite · FriendliAI · LMDeploy) or a customer-chosen cloud LLM (hybrid) | Local engine only |
| Identity | Email and password, or OIDC SSO; SCIM 2.0 | OIDC SSO, SCIM 2.0 | OIDC SSO (Okta, Keycloak, Authentik, Entra ID), SCIM 2.0, RBAC | Same, against the internal IdP |
| Time to first answer | Minutes (sign-up) | Days | About one week to a working POC | About one week where hardware exists |
| Support tier | Priority support, 99.9 % uptime | Dedicated support under an SLA, 99.9 % uptime | Production support contract (included in the recurring option) | Production support contract (included in the recurring option) |
| Pricing structure | $20/seat/month | from $2,000/month, flat, usage allowance included | from $10,000, one-time or recurring — no per-seat licence | from $10,000, one-time or recurring — no per-seat licence |
| Who it fits | Teams with non-sensitive or internal data wanting a start today | 50–2,000-staff firms that need someone else to run it | Regulated organisations with data-residency rules | Banks, government, defence: policy forbids any external network |
Self-hosted (community)
Free, MIT, unlimited users, community support, you operate it. One install command, Docker Compose, about five minutes.
There is no seat minimum for On-premises or Air-gapped. A single regulated team can run DocsGPT inside its own network under the same MIT licence, with the same connectors, agents, guardrails and identity integration as a two-thousand-seat rollout; pricing is flat infrastructure plus support, not a licence per user.
Some vendors publish on-premises only above several thousand seats — Langdock offers on-premises via Helm from 5,000 seats and nothing air-gapped; Glean has no true on-premises option. DocsGPT's On-premises and Air-gapped postures start at 1 seat.
Facts checked on
Per-seat products price every user, cap heavy users on tokens and add metered API overage. A self-hosted or on-premises DocsGPT deployment is flat infrastructure plus support, with no per-seat licence.
| List price | Seat minimum | On-premises | |
|---|---|---|---|
| Microsoft 365 Copilot | $30/user/month list | None published (requires M365) | No |
| Glean | ~$50–75/user/month (not published) | ~100 seats | No true on-premises |
| DocsGPT | $20/seat/month (Cloud); Managed from $2,000/month; On-premises and Air-gapped from $10,000, one-time or recurring | None | Yes, from one seat |
List prices and published minimums, checked 2026-09-06. Glean does not publish pricing; the range is as reported.
Single sign-on, provisioning, roles and an append-only audit table are in the open-source build and available in every posture.
SSO
Authorization Code + PKCE, group allowlists, admin-group mapping and back-channel logout against your identity provider.
AUTH_TYPE=oidc · Okta · Keycloak · Authentik · Entra ID
Provisioning
User lifecycle driven by the IdP: accounts are created, deactivated and reactivated from your directory. Deprovisioning revokes sessions immediately.
scim 2.0 · users · immediate revocation
Roles
User and admin roles across the workspace; Viewer / Editor sharing per agent and per source, applied at retrieval time.
admin · user · viewer · editor
Audit
Authentication events are recorded append-only and can be exported for compliance review.
auth-events · append-only · exportable
Detail on roles, audit and analytics: Admin, Security & Analytics · Security.
Local inference runs on the customer's existing GPUs. The bank reference runs its document-classification pipeline on NVIDIA T4 GPUs. Model choice is benchmarked against the target hardware before rollout, so the model fits the GPU rather than the other way round.
Hardware fit and sizing by model class →
Engines and registry: Models & BYOM.
Qwen3.8-Flash-Next
Ollama · http://ollama:11434
GLM-5.3-Flash
vLLM · http://gpu-01:8000/v1
GPT-5.6 Sol
DocsGPT · api key · via Arc53
+ Add model bring your own · openai-compatible
Six steps from scoping call to rollout.
30-minute scoping call
Posture, data sensitivity, seats and identity provider.
cloud · managed · on-premises · air-gapped
Posture and model choice
Hardware check where inference will run locally.
gpu class · vram · model candidates
Install
Cloud sign-up, Managed provisioning by Arc53, or the one-line installer (Docker Compose) or Kubernetes manifests in the customer's environment.
docs.ac/install · docker compose · deployment/k8s
Connect sources and the IdP
Two or three sources to start; OIDC single sign-on and SCIM.
sharepoint · google drive · confluence · s3 · oidc
One-week POC
A benchmark suite is the acceptance test.
docsgpt-cli bench
Rollout
Teams, roles, audit and channels. Production-ready in 30 days.
teams · roles · audit · widget · telegram · discord
Cloud
Start today
Internal or non-sensitive data, and a team that wants to start today.
$20/seat/month · minutes to start →
Managed
Someone else runs it
50–2,000 staff, client-confidential data, no platform team to run it.
from $2,000/month · flat →
On-premises
Your data centre
Data-residency rules; your data centre or private cloud; hybrid LLM optional.
from $10,000 · one-time or recurring · one-week POC →
Air-gapped
Zero egress
Policy forbids any external network. Local LLM, embeddings, storage; proven in a bank on T4 GPUs.
zero egress · from $10,000 · from one seat →
A 30-minute demo on your use case, or a free on-premises review of your hardware and identity setup.