jinni
Jinni generates a consolidated text dump of a project's relevant files (with per-file path headers) to provide LLMs with project context. It includes an MCP server exposing a read_context tool for agent-driven context retrieval with .gitignore/.contextfiles-style filtering, and a CLI that outputs or copies the context (or just lists included file paths) for manual use.
Score Breakdown
⚙ Agent Friendliness
🔒 Security
Primary risk is local filesystem exposure: the tool reads project files and can be constrained via MCP project_root/targets and an optional server --root for safety. No auth is described (since it’s local); thus the security model relies on restricting filesystem scope and filtering. Output can include large amounts of source text; ensure you don’t inadvertently include sensitive files by tightening ignore rules and roots. Dependency hygiene cannot be fully verified from provided content; dependencies include mcp, pathspec, pyperclip, pydantic, tiktoken.
⚡ Reliability
Best When
You’re using an MCP-enabled developer agent/IDE and want an automated, filterable snapshot of local repo files as LLM context.
Avoid When
You cannot safely constrain file access (e.g., via an explicit --root) or you need incremental/streaming retrieval instead of a single context dump.
Use Cases
- • Give an LLM a targeted view of a codebase (single module, directory, or the whole repo) for debugging or feature work
- • Automate “project context” retrieval in MCP-capable IDE/agent clients (Cursor, Cline, Roo, Claude Desktop, etc.)
- • Pre-generate context dumps for copy/paste into systems that don’t integrate with MCP
- • Reduce prompt/token waste by excluding binaries, dotfiles/hidden dirs, logs/build/temp files using ignore-style rules
- • Control context size with a configurable 100MB default and get actionable error details when exceeded
Not For
- • Production-grade secure multi-tenant deployments without additional isolation (it reads local filesystem content)
- • Situations where the LLM should not be allowed to access arbitrary files within an allowed root directory
- • Long-running interactive workflows that require streaming or pagination (it returns a single concatenated string)
Interface
Authentication
Authentication/authorization are not described in the README; the primary control mechanism is filesystem scoping via project_root/targets and optionally constraining reads with a server --root argument.
Pricing
No pricing information provided; appears to be an open-source tool distributed via PyPI/uv.
Agent Metadata
Known Gotchas
- ⚠ Cursor may silently drop context if it exceeds client maximum size; reduce what you request (e.g., narrower targets).
- ⚠ If you pass empty rules list [] in MCP, built-in defaults are used; if non-empty, built-in defaults and .contextfiles are ignored (easy to misunderstand).
- ⚠ Targets must resolve inside project_root; providing targets outside may fail.
- ⚠ Context size can abort with DetailedContextSizeError; agents may need to retry with smaller targets or tighter rules.
Alternatives
Full Evaluation Report
Comprehensive deep-dive: security analysis, reliability audit, agent experience review, cost modeling, competitive positioning, and improvement roadmap for jinni.
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.