Project Lex
Your SDLC, orchestrated.
The orchestration layer for enterprise engineering teams. Dynamic agents, enforceable policies, full audit trails - coordinated under one governed control plane.
Refactor auth module to PKCE flow, add tests, open a PR
Auth module refactored to PKCE. Dispatched researcher-codebase for context, then executor-code for implementation.
One set of rules.
Every team follows them.
Define how AI operates in your organization - who can do what, which models to use, what gets logged, and when humans approve. Then let every team and every agent inherit it.
Sandboxed Execution
Agents write and run code in isolated sandboxes. Every action is validated before it touches your codebase.
Codebase Awareness
Every agent understands your codebase - functions, types, dependencies, and relationships.
Autonomous Agents
Deploy agents into CI/CD that follow your team's standards. PR reviews, security checks, and deployments - governed by your rules.
Complete Visibility
Every agent action, every cost, every policy decision - logged and queryable across your organization.
Declarative Workflows
Define multi-stage workflows once. Every team inherits the same review, test, and deploy process.
Parallel Work, Zero Conflicts
Multiple engineers and agents work on the same repo simultaneously - each in an isolated branch.
Institutional Knowledge
Decisions, patterns, and context persist across sessions. New team members inherit organizational knowledge.
Runtime Policies
Set boundaries on file access, network calls, and spend. Policies are enforced at runtime - agents can't bypass them.
Full visibility.
Total control.
See what every agent does, what it costs, and whether it followed the rules. One dashboard for engineering, finance, and compliance.
Define your agents.
Researchers, planners, executors, reviewers - each with their own model, tool access, and guardrails. Create custom archetypes or use platform defaults.
Lex, Research, Plan, Execute, Review
Assign the right model to each role
Control file, network, and shell access
Track every change with full rollback
Know what AI costs.
Spend by model, by team, by project. Agent run volume, token usage, cost-saving from complexity routing - all in one dashboard your CFO can read.
Thresholds, hard caps, and notifications
Cost attribution by team and project
AI-generated cost reduction recommendations
CSV and API for finance tooling
Engineering intelligence.
DORA metrics, developer experience signals, and agent impact analysis across your engineering organization. See how AI accelerates your teams.
Deployment frequency, lead time, MTTR
Measure AI contribution to velocity
Developer satisfaction and friction points
Compare performance across squads
Every action, audited.
Who ran what, when, and at what cost. Policy violations flagged in real time. Export-ready for SOC 2, ISO 27001, and internal compliance reviews.
Every tool call, file edit, and API request
Flagged and categorized in real time
SOC 2, ISO 27001, custom formats
Filter by user, agent, action, or time range
Real-time observability.
Monitor every session, every cost, every model call. Spot anomalies before they become incidents. One dashboard for engineering and finance.
Watch active agent runs in real time
Spike alerts for cost and error rate
Latency, token usage, and error rates
One dashboard for eng, finance, and ops
Design pipelines visually.
Drag agents, checkpoints, and logic gates onto a canvas. Connect them into multi-stage workflows with parallel tracks, approval gates, and automatic validation.
Task, checkpoint, logic, start, and end nodes
Fan-out to multiple agents, fan-in at checkpoints
AI-powered graph validation before execution
Step through workflow execution visually
Explore the Control Plane
A fully interactive demo dashboard with realistic data. No sign-up required.
Your workflow.
Your environment.
VS Code extension, terminal CLI, or desktop app. Every client runs the same orchestration engine with the same policies. Pick the interface that fits - governance follows.
The editor, augmented.
Multi-agent execution inside VS Code, backed by the same routing, policy, indexing, and recovery systems as the full runtime.
Use Lex inside VS Code Chat, not just a custom sidebar
Hub view for orchestration, conversation panels for deep work
Spin up conversation-specific git worktrees when a task needs a clean lane
Review, keep, or undo generated changes before they land
Fan out specialists and watch each lane complete independently
Full tool timelines survive tab closes, restores, and panel reopen
Pull in files, selections, terminal output, and problems with token discipline
Reindex and search repo knowledge without leaving the editor
Route simple and heavy tasks to the right model tier automatically
Catch silent hangs from lost progress or lost communication
Turn staged diffs into commit messages directly from the extension
Health checks, auth status, model settings, and org policy refresh built in
LEX - Your SDLC, orchestrated.
Model: claude-sonnet-4-20250514 Status: ● Ready User: jamie.l Org: acme-corp
Welcome to Lex - Your SDLC, orchestrated.
/model • Switch provider, model, or agent
/conv • Browse & switch conversations
/new • Start a new conversation
/help • Show all commands
✦ Lex → dispatching workflow code-quality-audit
├ Stage 1 scan ········ ✓ 3 subagents, 847 files
├ Stage 2 analyze ····· ✓ 12 findings
└ Stage 3 report ······ ⧖ generating...
The terminal, unleashed.
Full agentic runtime in your terminal. Same engine, same policies, no GUI required.
Model switching, conversations, history
Pick up where you left off across restarts
Run workflows unattended in CI/CD pipelines
Pipe output to scripts and other tools
Lex, Research, Plan, Execute, Review
Right model for the right task, automatically
Projects
🛠 Tooling
⚙ Settings
❆ Lex → dispatching feature-implementation
├ Research ········ ✓ analyzed 23 files
├ Plan ··········· ✓ 4-step migration
├ Implement ···· ⧖ modifying 6 files...
Msrc/auth/session.ts
Msrc/auth/middleware.ts
Asrc/auth/jwt-provider.ts
3 files · +84 -12
@@ session.ts @@
- const token = crypto.random()
+ const token = jwt.sign(payload)
- res.cookie('sid', token)
+ res.cookie('jwt', token)
The desktop, refined.
Native desktop app with the same orchestration engine. Offline-capable, multi-project workspaces, and a built-in review pane for staging diffs before they land.
Organize conversations by project folder
Stage, revert, and commit diffs without leaving the app
Full shell with xterm.js, resizable drawer
Pick up where you left off across restarts
Same policies, models, and audit trails as the extension
macOS, Windows, and Linux via Electron
Built for teams
that ship at scale.
A multi-platform, provider-agnostic LLM client for VS Code, CLI, and Desktop. Define your rules once - every team inherits them.
Essential
Full platformEvery agent archetype, every tool, every harness. Nothing artificially gated.
- All agent archetypes & tools
- Accelerated engineer onboarding
- Multi-threaded workflows
- Persistent cross-session memory
- Codemode (sandboxed execution)
- Git worktree isolation
- SSO & BYOK encryption
- Control plane dashboard
Standard
Governance depthCustom governance, intelligent model routing, and deep observability for scaling teams.
- Everything in Essential
- Shared knowledge sources
- Organization workflows & skills
- Runtime policy engine & hooks
- Complexity-driven model routing
- MCP server integrations
- External service connections
- Session monitoring & replay
Enterprise
Full autonomyAutomated code review, CI/CD agents, compliance exports, and dedicated support for regulated environments.
- Everything in Standard
- Automated code review
- CI/CD integration agents
- Multi-environment deployment
- Cost intelligence & budgets
- SOC 2 & ISO 27001 exports
- Custom model routing rules
- SLA guarantees
Platform fee + per-seat licensing · Annual agreements