roampal-core
roampal-core is a Python MCP server plus a local HTTP hook/service that injects relevant long-term and outcome-scored memory into AI coding assistants (Claude Code and OpenCode). It also scores prior exchanges, promotes good advice, demotes bad advice, and supports ingesting reference documents for retrieval via semantic search across multiple memory collections.
Score Breakdown
⚙ Agent Friendliness
🔒 Security
The README describes a local HTTP server with a localhost health check but does not document TLS requirements, authentication/authorization, or secret-handling/logging behavior for HTTP/MCP endpoints. Scoring models may be remote (Zen free models by default per README), which can affect data exposure depending on configuration.
⚡ Reliability
Best When
You want fully local/private persistent memory for AI coding assistants and can run a small local service (and optional sidecar scoring model) on your machine.
Avoid When
You require strict enterprise-grade authz/authn, audit logging, and documented data handling controls for remote or multi-user deployments.
Use Cases
- • Persistent memory for AI coding assistants across editor sessions
- • Capturing and learning from outcomes of prior advice to improve future responses
- • Semantic retrieval over ingested documents (books) and historical conversations
- • Maintaining durable coding preferences and identity facts (memory_bank)
- • Automated context injection into Claude Code (hooks) and OpenCode (plugin)
Not For
- • Use as a public-facing multi-tenant SaaS API without additional hardening
- • Environments that cannot run a local HTTP server / local background scoring sidecar
- • High-assurance environments requiring formally documented security guarantees for all endpoints
Interface
Authentication
No explicit authentication mechanism is described in the README for the local HTTP server (example health check uses localhost). The MCP tools appear to be consumed by the host coding assistants configured during roampal init.
Pricing
Described as completely free and open source; costs may arise from using remote models for scoring if local models are not configured.
Agent Metadata
Known Gotchas
- ⚠ OpenCode scoring uses an independent sidecar; if unavailable, users must configure it (and the MCP tool score_memories is not registered on OpenCode).
- ⚠ The system depends on the local HTTP server being reachable at the configured port (27182 in examples).
- ⚠ If Claude Code hooks are not loaded, context injection may fail until Claude Code is restarted.
Alternatives
Full Evaluation Report
Comprehensive deep-dive: security analysis, reliability audit, agent experience review, cost modeling, competitive positioning, and improvement roadmap for roampal-core.
AI-powered analysis · PDF + markdown · Delivered within 30 minutes
Package Brief
Quick verdict, integration guide, cost projections, gotchas with workarounds, and alternatives comparison.
Delivered within 10 minutes
Score Monitoring
Get alerted when this package's AF, security, or reliability scores change significantly. Stay ahead of regressions.
Continuous monitoring
Scores are editorial opinions as of 2026-03-30.