Skip to content
Agentmart
Submit a tool

Gnt

Git-native policy layer for AI agents: check_action verdicts against rules approved via PR.

Connector by gnt-ai · GitHub account verified by the MCP Registry

IndexedApache-2.0Remote (HTTP)★ 30
Write a review

Install

Works with Claude Code, VS Code, Cursor, Windsurf. Read the repository before you give any tool access to your data.

npx agentmarket install connector/gnt-ai-gnt

Asks which of your tools to add it to (Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, Claude Desktop) and shows the change first. See https://agentmart.dev/cli.

Remote endpoint

  • https://api.gntai.dev/mcp Remote (HTTP)

Get the weekly diff

The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.

We use your email only for this newsletter. Privacy

About

Git-native policy layer for AI agents: check_action verdicts against rules approved via PR.

Use cases: identity-access, legal-compliance

Reviews

Write a review

No reviews yet. Installed it? Reviews come from verified installers only. How reviews work

Permissions

What this server can do and what it asks for, from its tool list, install config and code. Checked Oct 25, 2026.

Tools
6 read · 0 write · 0 deleteRead-only
Secrets
None asked for
Install scripts
Nothing to install (hosted)
Runs as
Hosted (remote endpoint)

Tools by what they do 6

Read from the source code; the running server may list different tools.

Read 6 · Look things up without changing anything.

  • check_action
  • check_company_policy
  • get_rule
  • get_skill_pack
  • list_skill_packs
  • search_rules

Network hosts it talks to 31

  • ai-gateway.vercel.sh · found in code
  • airtable.com · found in code
  • api.airtable.com · found in code
  • api.figma.com · found in code
  • api.github.com · found in code
  • api.gntai.dev · found in remote endpoint, install command, code
  • api.hubapi.com · found in code
  • api.intercom.io · found in code
  • api.linear.app · found in code
  • api.resend.com · found in code
  • api.zeroentropy.dev · found in code
  • astral.sh · found in code
  • bun.sh · found in code
  • docs.astral.sh · found in code
  • docs.docker.com · found in code
  • github.com · found in code
  • gitlab.com · found in code
  • gntai.dev · found in code
  • hermes-agent.nousresearch.com · found in code
  • linear.app · found in code
  • mcp.atlassian.com · found in code
  • mcp.granola.ai · found in code
  • mcp.linear.app · found in code
  • mcp.notion.com · found in code
  • mcp.zoom.us · found in code
  • ollama.com · found in code
  • railway.com · found in code
  • slack.com · found in code
  • spacy.io · found in code
  • turbo.build · found in code
Show 1 more hosts
  • yourteam.atlassian.net · found in code

Files it touches

  • Your home directory (~)

Tools (6)

Read from the source code; the running server may differ. Descriptions are the server’s own words; hidden characters are shown as [U+…].

  • check_action

    Check whether a described action complies with this org's approved, human-reviewed rules BEFORE the agent takes it. Returns a verdict — 'allowed', 'blocked', or 'needs_human' — with the specific rule(s) cited and a one-line reason. Call this before any side-effectful or hard-to-reverse action (sending a message, issuing a refund, deleting data, spending money). Treat 'blocked' as a stop, and 'needs_human' as "escalate, don't proceed on your own". Conservative by design: if no approved rule covers the action, retrieval fails, or the check can't be completed, the verdict is 'needs_human' — it …

  • check_company_policy

    Check the company's actual approved policy before acting on a billing or refund request. Returns a grounded verdict: allowed, blocked, or needs_human, citing the specific rule it's based on.

  • get_rule

    Fetch a single approved rule by id, with its provenance footer. Raises if the rule doesn't exist, isn't approved, or belongs to a different org — never leaks which of those is true. Accepts the pre-rename `id` param name too, for callers not yet updated to `rule_id` — `rule_id` wins if both are given.

  • get_skill_pack

    Fetch a compiled skill pack's manifest and file list (paths + hashes, not full file content) by id.

  • list_skill_packs

    List every compiled skill pack version for this org, newest first.

  • search_rules

    Semantic search over this org's approved, human-reviewed rules, optionally filtered to rules carrying any of the given tags. Only status == 'approved' rules are ever returned — draft, in_review, and deprecated rules are never exposed to a calling agent. An empty list is the unambiguous "no approved rule covers this query" signal — every returned hit already cleared the relevance threshold, so there's no separate "found something, but it's thin" case to represent. That gap is logged for the org (see `gnt gaps`).

Signals

Stars
30
Forks
29
Open issues
35
Contributors
22
Last commit
Oct 22, 2026 (6 days ago)
Last release
Sep 4, 2026 (1 month ago)

From GitHub (fetched Oct 24, 2026).

Details

Registry name
io.github.gnt-ai/gnt
Version
0.6.0
Auth
Auth not stated
License
Apache-2.0
Repository
github.com/gnt-ai/gnt
Website
gntai.dev
First seen
Oct 24, 2026
Last checked
Oct 24, 2026
Source
Official MCP Registrygithub-graphql
Cite this page

Gnt MCP server. Agent Mart. https://agentmart.dev/connectors/gnt-ai-gnt (last updated 2026-10-24)