正在加载视频...

视频加载失败

Introducing the Knowledge Base MCP. Greptile maintains a detailed knowledge base of how every part of your codebase works, as well as a history of past incidents. Starting today coding agents like Codex, Claude can access the Knowledge Base via Greptile's MCP, so they can use Greptile's learnings to...

64,850 次观看 • 9 天前 •via X (Twitter)

22 条评论

Greptile 的头像
Greptile9 天前

Learn more:

Simantak 的头像
Simantak9 天前

Yalls dont stop shipping do you. Love it team

Max 的头像
Max9 天前

Greptile knowledge base is slept on. It's so powerful having all the context in one place & constantly evolving

Martin Jonsson 的头像
Martin Jonsson9 天前

Does it work on historical? Lots of reviews behind me.

Saai Arora 的头像
Saai Arora9 天前

Damnnn this is fire

Varik Verilion 的头像
Varik Verilion9 天前

Incident history inside the coding loop is a useful addition.

EKOS _ AGI 🦊 🇮🇷 的头像
EKOS _ AGI 🦊 🇮🇷8 天前

This is where agent memory starts becoming real infrastructure. The next challenge is making that accumulated knowledge structured, versioned, and traceable — so agents don’t just remember what happened, but understand how that knowledge evolved and where it came from.

Yijie 的头像
Yijie8 天前

production quality on this is crazy

cole murray 的头像
cole murray9 天前

awesome!

Samuel Hu 的头像
Samuel Hu9 天前

the history-of-incidents bit is what i actually want. most mcp demos can answer a file question, but the "why did we do this weird thing" context is the part i keep missing — does it update itself from prs or only docs?

Dominik 的头像
Dominik9 天前

the useful boundary is not 'agent can read the codebase.' it is whether each retrieved fact carries provenance, freshness, and an owner. otherwise a knowledge base becomes another prompt-shaped authority with no way to tell stale guidance from a current constraint.

AWE 的头像
AWE8 天前

does the incident history get pulled in automatically or does someone still have to write the postmortem

Ali Mirza | AI Agents & Automation 的头像
Ali Mirza | AI Agents & Automation8 天前

hooking an mcp incident log directly into the subagent prompt stops it from making the same breaking imports twice.

Bessi 的头像
Bessi8 天前

past incidents in the MCP. did not expect that

virtualuncle 的头像
virtualuncle8 天前

most codebase context tools tell the agent how things work. knowing how it broke last time is a step past that

Mike Lopez 的头像
Mike Lopez8 天前

How do you prevent stale architectural knowledge from becoming authoritative context for the agent?

Mohamed Khlifa 的头像
Mohamed Khlifa8 天前

What is the difference between it and things like onsidian git that is free and most common for users ?.

0xdotdot 的头像
0xdotdot8 天前

fire

Omair-Inam Abdul-Matin 的头像
Omair-Inam Abdul-Matin9 天前

How much will this cost?

Jag 的头像
Jag9 天前

Interesting. So all the code base replicates to greptile?

AI Mastery Guide 的头像
AI Mastery Guide8 天前

past incidents remembered, that's smart

TonkaToyXL | AI Test Cabin 的头像
TonkaToyXL | AI Test Cabin8 天前

This is the missing piece for coding agents. Giving them a living map of the codebase and past incidents is how they stop guessing and start helping for real.

相关视频