794
total weekly downloads
11 packages
@opafex/create-mcp
Scaffold a new MCP server project with @opafex/mcp-shared
@opafex/grounding
Deterministic checks that an LLM agent is backed by its tool results: fabricated identifiers, citation drift, self-contradiction, unsupported counts, unsupported money, percentages, counts, dates and reference numbers, invented written values, and - befor
@opafex/knowledge-grid
Structured context for AI agents — a layered knowledge grid with vector indexing, domain inference, intent-driven retrieval, and token-budgeted prompt composition. Better context, better agent performance. Zero dependencies.
@opafex/mcp-approval-queue
Human-in-the-loop approval gate for AI agents: risk scoring, decisions in the client's own UI via MCP elicitation, and a CLI. The agent can ask; only a human can approve.
@opafex/mcp-enrichment
MCP server for multi-source entity enrichment (company, contact, email, phone)
@opafex/mcp-file-processor
MCP server for universal text extraction, keyword extraction, language detection, and text chunking for RAG pipelines
@opafex/mcp-shared
Shared utilities for building MCP servers — server factory, caching, config, logging, validation
@opafex/mcp-web-research
MCP server for web research — search, SSRF-safe page fetching, structured extraction and comparison
@opafex/privacy
PII redaction for logs and LLM payloads: same value, same token, so a redacted log is still debuggable. UK-aware, zero dependencies.
@opafex/relay
Let a hosted opafex agent use the commands and local MCP servers you allow on your own machine. Outbound only, no shell, nothing off your list.
@opafex/retrieval-core
Storage-agnostic retrieval for RAG: chunking, IDF-weighted rare-term matching and a vector + lexical merge, over three small ports. Zero dependencies.