Product workflow
Capture conversations, recover project state, assemble context, ask across workspaces, and review proposed changes.
Resources
Deployment, MCP, and security notes.
Public-facing reference material for evaluating ReasonTrail. Detailed installation commands should live in the customer/admin documentation, not on the marketing page.
Self-hosted deployment
Run the application and database on infrastructure controlled by the customer or operator. Keep public exposure disabled until hardening is complete.
MCP connector
Connect coding assistants to project memory so work can be checked against existing context and written back as reviewable records.
Security model
Local authentication, server-side sessions, database backups, export paths, AI provider boundaries, and explicit review before important writes.
Self-hosted deployment model
ApplicationSelf-hosted web app runtime
DatabasePostgreSQL controlled by the deployment operator
AuthenticationLocal email/password accounts with server-side sessions
AI providersOptional. Selected context may be sent to the configured provider or local model.
ExportsJSON and Markdown export paths for portability
BackupsDatabase backup policy owned by the deployment operator
MCP connector model
Use caseLet Codex, Claude Code, or another MCP-capable assistant read relevant project context before work starts.
Write pathAssistant writes should create reviewable captures, tasks, decisions, or links rather than silently accepted project facts.
AccessKeep MCP endpoints private. Use authentication or a private-network boundary.
Default behaviorIf a conversation has no known project, store it in an inbox or default workspace until reviewed.
Public launch readiness
The public site should stay factual. Claims about compliance, hosted availability, certifications, and security controls should appear only when they are implemented and documented.
Document the operator identity before public launch.
Publish privacy, cookie, terms, and security pages before collecting production customer data.
Keep self-hosted deployments on private networks until hardening is complete.
Separate customer-controlled data from provider-submitted AI context in all disclosures.