MCP server for Claude Code & Codex

Your coding agent has no idea why.

It writes the code in seconds, then breaks the thing the team decided last week, because it can't see Linear, Slack, the incident, or the sync notes. Hog feeds your agent that context over MCP, before it writes a line.

Try the live demo →

Works with Claude Code & Codex · two-tool MCP server · runs on your data

Pulls context from the tools you already use

"Why did the agent do that?"

Because it didn't know ENG-102 was blocked by the Inngest incident. It never read the eng-sync note where you decided to slip the launch. So it built on assumptions, you caught it in review, and you lost the afternoon.

The agent isn't dumb. It's blind. Every prompt starts from zero context about your product.

The fix

Two tools. Instant grounding.

hog_search and hog_ask drop into any MCP agent. It asks "why is Checkout v2 delayed?" and gets the cited answer, pulled across every tool, before it touches a line of code.

✓Cited, not guessed. Every answer traces back to the Linear ticket, Slack thread, note, or incident it came from.
✓Decisions, not just diffs. The agent learns why the team chose something, not only what changed.
✓One install. Point your MCP config at Hog. No new workflow for the engineer.

claude-code · hog (MCP)

# agent needs context before editing checkout

▸ hog_ask("why is Checkout v2 delayed?")

Launch slipped to next sprint. Root cause:

ENG-102: Inngest webhooks failing, traced to

the Inngest incident on 2026-06-22. Decision to

slip was made in the eng sync notes.

# sources: Linear · Slack · Notes · Status

▸ agent now edits with full context ✓

How it works

Three steps. No new workflow.

Hog runs alongside your agent and answers over MCP.

Connect your tools

Linear, Slack, notes, and live status connect read-only. Any tool with an API.

→

Hog links them

Refs, mentions, authorship, and cause-and-effect become one queryable graph.

→

Your agent queries it

Claude Code and Codex call hog_ask over MCP and get cited context.

Where it goes

Start with your agent. Grow into your team.

The same graph that grounds your coding agent answers your whole team's "why did we decide this?" The agent is the way in. The product context layer is where it goes.