Enclave sessions
Run interactive sessions inside the enclave. Analyse data, build workflows and query either model, with nothing leaving in plaintext.
Sovereign intelligence
IntelligenceWithoutPermission.
Inlex AI is a decentralized AI network that runs every request inside a sealed hardware enclave. Your prompts and your keys stay under your control, and the operator running the hardware has no path to any of them. Two open weight models are available: Shard for low latency work at the edge, and Axis for full capability reasoning under a policy you write yourself.
- Execution
- Sealed enclave
- Telemetry
- None retained
- Key custody
- Held by you
- Context window
- 256K tokens
Connects to
SlackNotionGmailGoogle DriveGoogle CalendarPostgresSnowflakeS3GitHub
01 / Execution
Private execution
Agents that never expose your data
Agents coordinate work across your stack from inside a sealed hardware enclave. Inputs are decrypted only within that boundary, the host operating system holds no readable copy, and the operator running the machine cannot inspect a session while it is live or after it ends.
Encrypted workspaces
Teams work in shared spaces where the host sees ciphertext and nothing else. Reports are compiled and insights are surfaced without proprietary material leaving the boundary in a readable form.
Secure integration grid
Enterprise data moves between cloud environments, local instances and air gapped systems along routes that keep the security boundary intact at every hop.
Signal from raw data
Fragmented data lakes are turned into operational answers inside the enclave. The underlying records are read for the query and are never copied into a third party context.
Policy enforced in hardware
Independent agents execute processes across your systems while the enclave enforces the policy you set. Compliance is demonstrated with an attestation record rather than a written assurance.
02 / Models
Shard and Axis
Open weights, running under your rules
Both models ship as open weights you can hold, audit and run. Shard is the lightweight model for edge and enclave execution where latency is the constraint. Axis is the full capability model, released without vendor imposed filters so that the usage policy is yours to write and yours to enforce.
Shard
Fast edgeBuilt for interactive work and high volume routing, with response times that hold up on a laptop enclave or a regional node.
- Parameters
- 9B
- Context
- 256K
- First token
- 90 ms
- Licence
- Open weights
Axis
Full capabilityBuilt for long reasoning, research and analysis, with no filter layer between the model and the policy your organisation defines.
- Parameters
- 27B
- Context
- 256K
- First token
- 310 ms
- Licence
- Open weights
03 / Product
Inlex AI in motion
Four surfaces, one boundary
Every request in the product takes the same route through a sealed enclave, whether it comes from the console, a connected database or a scheduled agent. The boundary does not change shape depending on which surface you are using.
Console
Integrations
Connected systems
Attach databases, enterprise APIs, warehouses and on chain systems. Credentials are sealed at connection time and used only inside the boundary.
Analytics
Team signals
Read organisation wide signals such as workload risk, blocked handoffs and decision latency, computed without exposing the private content behind them.
Automation
Autonomous workflows
Deploy agent loops that adapt to changing inputs and run on a schedule, isolated from any third party observer including the node operator.
04 / Security
Encrypted by design
Privacy and speed on the same path
Inlex AI runs on hardware sealed enclaves with attestation on every session, which removes the usual trade between confidentiality and response time. There is no privacy mode to switch on, because there is no other mode.
Cryptographically sealed
Each run happens inside a certified enclave. Access credentials and decryption keys stay in your possession for the whole lifetime of the session.
Private global routing
Requests cross a global network without leaving a routing footprint that would reveal system layout, team structure or data location.
Keys that rotate
Keys rotate every 30 days, or immediately on device loss. Threads are resealed under the new key without writing plaintext at any point.
session sealed region EU-WEST-04 model axis-27b enclave TDX · quote verified measure 0x9f31c8a4be07d2... host read denied telemetry none retained latency 14:02 · 1.9s keys rotate in 21d
05 / Network
$INLEX
The fuel of the network
Every job on Inlex AI is priced in dollars and settled in $INLEX. Subscribers never handle crypto, because their payment converts into the tokens their agents burn. Node operators take 60 per cent of revenue for running attested hardware, 20 per cent funds the company, and 20 per cent buys $INLEX on the open market each month.
01Customer pays in dollars
02Operators run attested enclaves
03Compute cost settles in $INLEX
04Buybacks execute monthly
05Usage locks circulating supply
60%Paid to the node operators who run the enclave hardware.
20%Funds engineering, security review and network operations.
20%Buys $INLEX on the open market, executed every month.
06 / Pricing
Plans
Pricing for secure scale
Every tier includes hardware isolated enclaves and the full agent toolset. What changes between them is allowance, routing priority and the level of dedicated hardware behind your workload.
Starter
Freeper month
- Private sessions inside a sealed enclave
- 256K context on every request
- Standard daily allowance on Shard
- Enclave isolation with zero trust verification
- Access to the standard agent library
- Up to three secure integrations
- Pay in $INLEX for 10 per cent off
Pro
$10per month
- Ten times the Starter daily allowance
- Priority routing across the inference grid
- Shard and Axis, with model routing
- MCP support for the tools you already use
- Prepaid 30 day periods, unused days carry over
- Attested execution with verification proofs
- Priority support, 24 hours a day
Enterprise
Customper month
- Daily limits set to your workload
- Admin dashboard with 10 seats included
- Pinned model versions with reproducible weights
- Dedicated enclave clusters with auditable attestations
- Unlimited hardened system integrations
- Enterprise SLA and a named technical contact
- Data residency and custom compliance policy
- Full API access for private cloud deployment
07 / Roadmap
Roadmap
From one console to a private compute mesh
The work runs in three phases. The first built the enclave backend and the preview console, the second is building the product around it, and the third extends the network across regions and chains.
Phase one · complete
- Architected the enclave first backend and secure API fabric
- Deployed enclave backed infrastructure and the core agent APIs
- Curated the training corpus and stood up model infrastructure
- Established zero leak telemetry and feedback handling
- Completed a closed preview with eight enterprise partners
- Launched the preview console with hardware enforced privacy
- Tuned inference pipelines for low latency enclave responses
- Standardised hardware attestations and compliance proofs
Phase two · in progress
- Launch the unified dashboard and encrypted control surfaces
- Expand the early access pool and gather production feedback
- Ship zero knowledge personalisation that keeps context private
- Release mobile clients that tunnel into a live enclave
- Establish data partnerships using zero knowledge queries
- Release multi agent orchestration for enterprise operations
- Add voice input transcribed entirely inside the enclave
- Publish the developer SDKs for custom enclave modules
Phase three · planned
- Open regional enclave hubs with latency guarantees
- Deploy federated architectures for edge execution
- Combine voice, private memory and local first execution
- Run developer outreach and technical certification
- Bridge to further chains for borderless settlement
- Launch the enclave marketplace for third party agents
- Begin research on private predictive models for regulated sectors
- Launch the Inlex Sovereign Cloud, an encrypted compute mesh
08 / Questions
Frequently asked
Questions worth a direct answer
Is my data private when I use Inlex AI?
Yes. Requests are decrypted only inside a sealed hardware enclave, and the enclave publishes an attestation quote that you can check against the published measurement before you send anything. The node operator, the network and Inlex AI have no read path into a running session.
Does Inlex AI train on my data or store it?
No. Prompts and responses are decrypted only inside the enclave, are never written to disk outside it, and are not used for training. When a session ends there is nothing left for anyone to hand over.
Can I connect the tools I already use?
Yes. Databases, warehouses, enterprise APIs and on chain systems connect through the integration grid, and MCP support on the Pro tier covers most existing tooling. Credentials are sealed at connection time.
What separates this from a standard AI assistant?
A standard assistant asks you to trust a policy document. Inlex AI gives you an attestation record, open weights you can hold, and keys that never leave your custody, so the privacy claim can be checked rather than accepted.
Can I run Inlex AI on my own hardware?
Yes. Both models are open weights and run on premise or fully air gapped. Enterprise deployments can use dedicated enclave clusters with hardware attestations you audit yourself.
Is this built for teams or for individuals?
Both. The Starter and Pro tiers are sized for individuals and small teams, and Enterprise adds seat management, dedicated enclave clusters, data residency and an SLA.
How does $INLEX capture value?
Jobs are priced in dollars and settled in $INLEX, so usage creates continuous demand for the token. A further 20 per cent of revenue buys $INLEX on the open market every month, and tokens burned by agents leave circulation.
How is the company funded?
By subscription and usage revenue, with 20 per cent of that revenue retained to fund engineering, independent security review and network operations.
Get started
Private. Sovereign. Unmistakably Inlex.
Open a session with Shard or Axis and watch the attestation come back before your first token does. No telemetry, no vendor filter, and no operator with a view into your work.