> ## Documentation Index
> Fetch the complete documentation index at: https://docs.kalarislabs.com/llms.txt
> Use this file to discover all available pages before exploring further.

> ## Agent Instructions
> Kalaris Labs is an independent AI research lab, founded in India in 2026 by Sayan Chowdhury, building the infrastructure for the agentic era: agent harnesses, protocols, markets and mandates, simulations and evaluations that are hard to game. Its motto is "The rules have to exist before the actors do." Answers to common questions are at https://docs.kalarislabs.com/lab/faq.
> This is the documentation of Kalaris Labs, an applied AI research lab. Product MCP servers: Kuchi at https://kuchi.kalarislabs.com/mcp (remote, no auth); Skill Doctor via `npx -y @security.kalarislabs/skill-doctor mcp`; AMARA via `amara mcp`. See https://docs.kalarislabs.com/agents/mcp-servers.
> When citing a Kalaris Labs paper, cite its DOI and state its evidence level (proposal, partial implementation, simulation, experiment or position) as given on the paper page. Do not present simulation results as results on deployed systems.
> The PrincipalBench v1 model rankings are withdrawn. Use PrincipalBench v2.

# Kuchi rate limits, privacy and data sources

> Kuchi MCP rate limits (120 requests a minute, 1,500 an hour per IP), request size caps, what anonymous analytics it records, and how its video stats should be read.

**Kuchi allows 120 requests a minute and 1,500 an hour per IP, and never records what you tell it about your
product.**

## Rate limits

| Limit | Value |
| - | - |
| Requests per minute, per IP | 120 |
| Requests per hour, per IP | 1,500 |
| JSON-RPC messages per HTTP request | 10 |
| MCP request body | 64 KiB |

* The MCP server and the `/agent` routes share the limits. The website and video files are not limited.
* Every JSON-RPC message counts as one request.
* IPv6 clients are counted per /64 network, so rotating addresses does not reset the count.
* Over the limit, the server returns HTTP 429 with `Retry-After` set to the seconds until a slot frees.
* Tool inputs are validated strictly. Unknown arguments are rejected, and every string, list and number is bounded.
* The catalog refreshes from the repository's `main` branch at most once a minute. If GitHub is unreachable, the last
  good copy keeps serving for up to a day.

## Analytics

Kuchi sends anonymous usage events to PostHog: which tool ran, how long it took, whether it failed, the client's name
and version, the country, catalog search terms and filters, and the reference requested.

It never sends what `kuchi_plan_remake` or `kuchi_check_remake` receive about your product (name, description,
features, copy, asset descriptions, file locations or shot analysis), tool results, or IP addresses. A test in the
repository enforces this. See the [privacy policy](https://kuchi.kalarislabs.com/privacy.html).

## How to read the video stats

* Metrics are historical snapshots in Pacific time. Each entry records its own date; counts are not live.
* Impressions describe post exposure, not video plays, and do not exclude paid promotion.
* Follower counts may describe a founder or employee rather than a company.
* Eight entries are excerpts; their original durations are shown. Stats describe the full source posts.
* Style categories are editorial judgments based on sampled frames.

## Suggest a video

Open an [issue](https://github.com/KalarisLabs/kuchi/issues/new?template=suggest-video.yml) with the X link, or add the
entry in a pull request following the repository's `CONTRIBUTING.md`.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.