Organization repository on GitHub
neatroots: see, check and replay your AI agent's runs on your machine while you build it. Local-first, by Neatlogs.
- Forks
- 49
- Last pushed
- Updated Oct 2026
Repo stats from the GitHub API, cached Oct 2026.
Jev directory
CommunityJev Tools & CLIs
Neatroots
A local-first tool from Neatlogs for watching what your AI agent did: it ingests the traces your agent already sends, checks every run, and gives coding agents 23 MCP tools to read, explain, search, compare, and replay runs.
- Stars
- 74
- Language
- TypeScript
- License
- APACHE-2.0
- Reading time
- 3 min
- Published by
- Community
- Author
- neatlogs
- Last verified
- Oct 2026
- Languages
- TypeScript
- Capabilities
- Open sourceRuns locally
Highlights
- Runs on your laptop, reads traces your agent sends, catches what went wrong, and hands the evidence to your coding agent; nothing leaves unless you send it.
- Takes runs from the Neatlogs SDKs (Python, TypeScript, Go, Browser), the HTTP trace API, any OTLP exporter over HTTP or gRPC, trace files, or Neatlogs Cloud.
- Checks every span with rules in code — loops, repeated tool calls, error streaks, leaked keys — plus yes/no questions each scored with a probability.
- neatroots connect gives Claude Code, Codex, Cursor and six others 23 MCP tools to read, explain, search, compare, run SQL, write checks, and replay runs.
- Runs locally with Ollama, LM Studio, llama.cpp, or MLX plus the local decision model Laya; GitHub: 73 stars, 47 forks, Apache-2.0, pushed 2026-10-17.
Quickstart
bash
neatroots connectWatch out
README reviewed 2026-10-17 and truncated mid-bullet in "Runs locally by default", so the cloud model options, install commands, and version requirements are not in the digest. The local decision model is Laya, linked to Hugging Face; the excerpt never mentions Jev, TypeSafe, or the Jev API.
Questions about this entry
Short answers drawn from the linked project or page.
A local-first observability tool where an AI agent starts out: it runs on your laptop, reads the traces your agent already sends, catches what went wrong, and hands the evidence to your coding agent, with data staying on the machine by default.
It picks yes/no questions for your agent and a decision model answers each with a probability, alongside rules in code that check spans for loops, repeated tool calls, error streaks, leaked keys, and failed or empty runs.
The digest excerpt names Laya, a local decision model linked to Hugging Face, and never mentions Jev, TypeSafe, or System One. Any Jev connection is by class resemblance, not documented integration.
The Neatlogs SDKs (Python, TypeScript, Go, Browser), the Neatlogs HTTP trace API, any OpenTelemetry exporter over OTLP HTTP or gRPC, trace-file imports, and production traces pulled from Neatlogs Cloud, with OpenInference, OpenLLMetry, GenAI, and Vercel AI SDK attributes.
neatroots connect gives Claude Code, Codex, Cursor and six others 23 MCP tools — read a run, explain it, search, compare runs, run SQL, write a check, replay — or you can chat with Claude Code and Codex inside the app with those tools attached.
Yes: re-run one LLM call with an edited message, model, or system prompt without an agent, or re-run the whole agent from the app; editing a user message in the Overview forks the run.
It is Apache-2.0 with 73 stars and 47 forks on GitHub, last pushed 2026-10-17. The README's run command is neatroots connect; the digest does not include install steps or system requirements.
More Jev resources like this
Latitude is an open-source observability platform for AI agents: one line of telemetry captures traces, failing traces group into tracked signals, and fixes are dispatched to coding agents and replayed against real traces.
Jev Tools & CLIs#community#catalog-import#logicrw
Community
latitude-dev/latitude-llmTypeScript
latitude-dev/latitude-llmTypeScriptA local guard layer for AI coding agents: it evaluates delivered tool hooks against deterministic rules, blocks dangerous ones in milliseconds with no model in the decision path, and keeps a grep-able local activity log for auditing.
Jev Tools & CLIs#community#catalog-import#logicrw
A very WIP tool for jev-ification from Harry Solovay: lessons from coding sessions become approved Markdown traits, and a configurable decision model evaluates code one trait at a time while an MCP server handles group-aware approval.
Jev Agents & MCP#community#github#typescript
Community
harrysolovay/traits-pocTypeScript
harrysolovay/traits-pocTypeScriptA per-turn model router for the Pi coding agent: a small judge picks the cheap or the strong tier for each message, with multi-model failover and task-level orchestration, plus an optional Jev judge that answers with a calibrated probability.
Jev Tools & CLIs#community#github#router
A local desktop workbench for inspecting agent runs: it scans a Codex directory, imports snapshots into SQLite, and lets you browse timelines, tool output and source evidence offline. Jev analysis of the runs is planned but not implemented.
Jev Tools & CLIs#community#github#desktop
Community
kailbug/jevalTypeScript
kailbug/jevalTypeScriptA type-aware TypeScript linter that pins a public typescript-go compiler module and runs syntax- and type-aware rules in one AST pass per root file; its semantic mode reads changed files and asks TypeSafe Noul questions over bounded spans and candidates per policy.
Jev Tools & CLIs#community#github#typescript
From the Jev guides
Original write-ups on what this entry is for.
From the Jev AI model community
Posts from builders shipping with Jev right now.