AXLR / Docs

Documentation

AXLR is Underpass's default engine for agentic execution, supported by KMP memory and MADE procedures and extended through compatible MCP servers and OpenAI/Codex plugin packages. Start with the product contract for responsibilities and the exact capabilities of each interface.

Start and use🔗

TaskGuide
Build and launch the consoleGetting started
Use models, work modes, approvals, sessions and diagnosticsConsole
Look up every key of settings.json, mcp.json and credentials.jsonConfiguration reference
Understand driven and explicitly requested proceduresWorkflows
Install packages or connect MCP toolsPlugins and MCP
Resolve a visible failureTroubleshooting

Operational runbooks🔗

OutcomeRunbook
Take a task from context recovery to verified deliveryAgent workflow
Connect, verify and disconnect memoryKMP
Prepare identity, definitions and tracked executionMADE
Qualify, install and remove an extensionExtensions
Update engines or recover interrupted work and backupsRecovery and maintenance
Check service health and reconcile API operationsService operations
Keep documentation aligned with a code revisionDocumentation maintenance

Integration and deployment🔗

The HTTP/gRPC/MCP parity ledger maps all 18 service capabilities and their shared request, approval and event contracts.

InterfaceContract
HTTP serviceService API, OpenAPI v1
KubernetesHelm deployment
One-request processJSON worker
Go hostLibrary, MCP client
Implementation and distributionArchitecture, releasing

Documentation status and design record🔗

The current guides were reconciled with main at 559a1b8 on 4 October 2026. The audit records source evidence, corrections, validation and remaining product limits. Brand documents the visual assets.

The tool registry design, with its pre-forged tools catalog and workflow catalog, proposes a public catalog of installable tools and workflows; its reference entries live under catalog/. Documents under plans, design studies, superpowers specs and superpowers plans preserve dated design and implementation work. The service design, context research, 27B-class local model research, diagnostics and provenance explain decisions; they are not installation or current API references. Use the current guides above for operations.

From docs/index.md at 2d95f7e in underpass-ai/AXLR View on GitHub