Roadmap · updated 25 Oct 2026
What is built, what is next.
This page is generated from our roadmap file, which we keep in step with our working plan. Dates are planning targets and can move. "Done" means built and tested by us in development or on our test server; it does not mean generally available. Nothing on this roadmap is generally available yet.
- Done Built and tested by us
- In progress Being built now
- Planned On the plan, not started
IQ platform
The shared foundation under every product: identity, security gateway, per-firm data, metering, agents and hosting profiles.
-
Architecture and security foundations
In progress19 Oct 2026 to 30 Oct 2026
The architecture is decided and the security baseline is in place on our test server.
- Layered architecture approved: every layer independently hostable (Done)
- Security gateway: TLS 1.3 only, services that verify each other, post-quantum ready (Done)
- Every call between products carries the person's identity and expires within minutes (Done)
- One database per firm; a request not tied to its firm is refused (Done)
- Metering of every API call (Done)
- Browser sign-in with a second step: a security key or an authenticator app (Done)
- A plan for post-quantum encryption, drafted (Done)
- Public test address with a dynamic security scan baseline (In progress)
-
Service contracts and conformance
In progress2 Nov 2026 to 13 Nov 2026
Every product speaks to the others through versioned interfaces that are tested automatically and against our test server.
- Products connected through versioned, tested interfaces (Done)
- End-to-end checks across every product on our test server (Done)
- Remote command-line access for operators (Done)
- Load testing and measured capacity (Done)
- One explicit boundary for hosted and local models (In progress)
-
Control plane and agent tier
In progress16 Nov 2026 to 27 Nov 2026
Firms are set up from one place, and AI agents work only with the permissions of the person they act for.
- Firm set-up, entitlements and audit in one place (Done)
- AI agents limited to approved tools, steps and budgets (Done)
- Tool gateway: an agent never has more access than the person it acts for (In progress)
- Import service for data from existing systems (In progress)
- Retention review queue and legal holds, decided by the firm administrator (In progress)
- Observability without client content (Planned)
-
Hybrid sync and corpus foundations
In progress30 Nov 2026 to 11 Dec 2026
The groundwork for running a firm across its own servers and the cloud at once.
- Two-site sync proven in a fault-injecting simulator (In progress)
- Entitlement-filtered access to the global corpus (Planned)
- Document storage on your own disks or in cloud storage (Planned)
- Folder, ZIP and manifest import with staging and rollback (In progress)
- AI models running on your own server (Done)
-
Two-site sync and on-premises installer
Planned14 Dec 2026 to 25 Dec 2026
Two sites stay in step securely, and a Linux server becomes a working site from an installer.
- Live sync between two sites with conflict handling (Planned)
- On-premises installer for Linux servers (Planned)
- Signed releases with verifiable provenance (Planned)
- Firm automation with approvals and kill switches (Planned)
- Spending caps for every kind of usage (Planned)
-
Hybrid pilot and release gate
Planned4 Jan 2027 to 15 Jan 2027
A hybrid pilot with a fictitious firm, failure drills, and end-to-end checks as the release gate in every hosting set-up.
- Hybrid pilot across two sites (fictitious firm) (Planned)
- Failure drills: network partition, restore, key rotation (Planned)
- Point-in-time recovery drills per firm (Planned)
- Firm agent builder (Planned)
- Release checks in cloud, on-premises and hybrid set-ups (Planned)
- Remaining screens designed once the interfaces are stable (Planned)
LexIQ
Legal research and drafting with cited answers from your own sources, and from public law once the CodexIQ corpus is released.
About LexIQ-
Alpha 1
In progressNow
Running on our test server with fictitious firms.
- Cited answers with AGLC4 footnotes (Done)
- Answer formats: short answer, research memo, essay and problem-question analysis (Done)
- Workspaces and matters with permanent matter codes (Done)
- Drafting with practitioner sign-off before export (Done)
- Share a draft with a reviewer (Done)
- Confidentiality levels, including a mode where nothing leaves your server (Done)
- Confidential matters with warn, allow or block rules (Done)
- Encrypted backups (Done)
- Search available to the other products (Done)
- Per-firm encryption of client text (In progress)
- Fully local answers (until released, the local mode returns cited passages) (In progress)
-
Beta 1
PlannedDate not set
LexIQ on the shared platform, ready for invited firms.
- Served only through the platform gateway (Planned)
- Global corpus access filtered by entitlement (Planned)
- Research and drafting agents in the agent tier (Planned)
- Defences and tests against prompt injection (Planned)
- Events to PracticeIQ: matter opened, document filed, deadline suggested (Planned)
-
Release candidate 1 and version 1.0
PlannedDate not set
General availability follows a release candidate and the release gates.
- Redesigned user experience (Planned)
- Release gates passed in every hosting profile (Planned)
PracticeIQ
Practice management built around matters, linked to LexIQ.
About PracticeIQ-
M1: matters and people
DoneBuilt October 2026, ahead of its 25 Dec 2026 target
Matters, users and roles, linked to LexIQ.
- Matters with permanent codes, team access and archive (Done)
- Deadlines that a person confirms (Done)
- Each matter linked to LexIQ under the same matter code (Done)
- Matter import from an existing system (dry run, then apply) (Done)
- Each request runs on the firm's own database (Done)
- Users, roles and firm settings (Done)
-
M2: calendar and mail
In progress4 Jan 2027 to 15 Jan 2027
Microsoft 365 calendar and mail, linked to matters.
- Calendar month and week views (Done)
- Microsoft 365 calendar sync (read first) (In progress)
- Mail linked to matters by a person or by rules (In progress)
- Deadlines and reminders from the calendar (Planned)
-
M3: research, costs and disclosure
In progressAfter January 2027
The matter becomes the place where research, drafting and AI costs meet.
- Research and drafting page tied to a matter (Done)
- AI cost ledger per matter (Planned)
- Disclosure tracker (Planned)
- User experience design (In progress)
CodexIQ
A curated corpus of legislation, case law and commentary, and deep research for firms.
About CodexIQ-
Curation foundations
DoneOctober 2026
The review workflow that every source passes through before it joins the corpus.
- Candidate review with facets (jurisdiction, area of law, kind) and provenance (Done)
- Numbered corpus releases on an append-only record (Done)
- Firm consent recorded before a source joins a firm's corpus (Done)
- Separate access for our curators and for firms (Done)
-
M1: first discovery
Planned14 Dec 2026 to 25 Dec 2026
Discovery on one jurisdiction and one area of law, starting from a small sample with a cost estimate.
- Source discovery for one jurisdiction and area of law (Planned)
- Facet classification, always reviewed by a person (Planned)
- Signed, versioned corpus packages (Planned)
-
M2: firm deep research
Planned4 Jan 2027 to 15 Jan 2027
Firms find new sources from inside LexIQ and choose, with a recorded consent, what joins their corpus.
- "Find sources" in LexIQ with a consent record (Planned)
- Corpus packages imported on a second site (Planned)
AegisIQ
Agentic security monitoring that detects, contains and reports, and never counter-attacks.
About AegisIQ-
Foundations
DoneBuilt October 2026, ahead of its 12 Nov 2026 target
The foundations every detection is built on and measured against.
- Security events that never carry client content (Done)
- A tamper-evident record of events and incidents (Done)
- Sensors on the gateway and every service (Done)
- Detection rules measured against replayed attacks (Done)
- Operator tools (Done)
-
M1 Watch
In progressTarget 26 Nov 2026
See attacks on the test server in near real time, with alerts.
- Detection of brute force, scans, floods and token abuse (Done)
- Host intrusion and tenant-isolation probing signals (Done)
- Detection of misuse by agents (Done)
- Malware scanning of every uploaded file (In progress)
- Alerts (In progress)
-
M2 Contain
PlannedTarget 10 Dec 2026
Reversible, expiring containment on your own systems, with approvals and a kill switch.
- Ban, revoke, disable and quarantine actions that expire and reverse (Planned)
- Ownership check before any action (Planned)
- Approvals, allowlists and a kill switch, proven by safety tests (Planned)
-
M3 Know
PlannedTarget 24 Dec 2026
Public vulnerability and threat intelligence matched to what you run.
- Public advisory and threat feeds (Planned)
- An inventory of the software inside every service (Planned)
- Prioritised exposure list with patch clocks (Planned)
-
M4 Assist
In progressTarget 14 Jan 2027
A local model will explain and group events; it will never take action. Firm views come first.
- Model-assisted triage on a local model (Planned)
- Incident reports and evidence packs (Planned)
- Support for a firm's breach-response process (Planned)
- Per-firm views and digests (Done)
-
M5 Release-candidate ready
PlannedTarget 28 Jan 2027
Patch proposals for review, adversarial testing and hosting profiles.
- Patch proposals as reviewable changes (never merged automatically) (Planned)
- Adversarial testing of AegisIQ itself (Planned)
- Hosting profiles: same host, cloud, on-premises, air-gapped (Planned)
Want to hear when milestones land?
Leave your email for occasional product updates, or register your firm for the beta.