Platform

Ask Anything, Safely

• Governed and audited every conversation runs under your organisation's rules, with a full audit log of what was said and by whom.
• Safeguarded conversations are monitored for safeguarding concerns as they happen, with alerts sent directly to admins or safeguarding leads.
• Grounded in your own knowledge on local models, responses draw on your organisation's own documentation and policy, not general internet knowledge.

Know Whats Happening

• Safeguarding at a glance flagged conversations route straight to designated safeguarding leads, with the context they need to act, not just an alert.
• Full audit trail every interaction, every model used, every user, logged and searchable, ready for an internal review or an external audit.
• Usage and cost reporting see exactly what's being used, by whom, and what it costs, with spend caps you control rather than discover after the fact.

Agentic Development on Your Terms

• Plans, then acts a change request becomes a short plan you can approve, then multi-file edits, then a test run, with every step visible.
• Runs tools with permission shell commands, tests and builds execute only with your approval, and their output feeds the next step.
• Model-routed each request goes to whichever local or approved cloud model has the right capability, without developers managing that choice.
• Grounded in your codebase suggestions draw on your own repositories, standards and approved libraries, not general internet training data.
• Governed like everything else every action is logged and attributable; coding mode skips document search and passes tool calls straight through, and safeguarding observes rather than blocks.

AI system administration for Linux and Windows

• Setup "install PostgreSQL with nightly backups to this path", "set up nginx with a Let's Encrypt certificate for this domain", "create a service account that can only restart Moodle".
• Debugging "why is the disk full", "what is using port 443", "why did the backup fail last night". eLLA reads logs, event logs and system state, explains the cause and proposes the fix.
• Maintenance patching, log rotation, certificate renewals, user and firewall changes, scheduled and reported.
• Safe by design read-only checks run instantly; every change is reviewed by a second AI, needs your approval, and is recorded with its undo, or flagged before you approve if it cannot be undone.
• Private by design runs against local eLLM models, so hostnames, configuration and secrets never leave your network. Runs on Ubuntu (bash, locally or over SSH) and Windows 10/11 and Server 2019–2025 (PowerShell).

One Platform, Different Goals

• A teacher marking a stack of essays who needs a first pass, not a replacement for their judgement.
• A safeguarding lead who needs to know the moment a conversation raises a concern, not weeks later.
• A college support team answering the same enrolment and finance questions, hundreds of times a term.
• A council caseworker drafting responses to residents without waiting on a template.

• A council FOI officer triaging requests and flagging what shouldn't be released.
• An IT lead who needs proof, not promises, that student or citizen data never left the building.
• A developer inside a government department shipping code without their tooling calling out to the open internet.
• A department head who needs a straight answer on what AI cost this term and who used it.