Graphlit

Context engineering for developers. Ingest any content, extract what matters, give your AI the memory it needs. Share via MCP with Claude, ChatGPT, etc

[Graphlit @graphlit]

🔥 New @graphlit platform release!

In addition to our MCP server, we've released support for OpenAI GPT-4.5, Sonnet 3.7, and the new @MistralAI OCR model for document preparation.

https://t.co/7lI814BHZZ

[Kirk Marple @KirkMarple]

Things are getting wild.

Asked our @dossium agent to analyze its content ingestion timeline of the last month.

It called our 'query_help' tool to codegen a call to the internal SDK, and executed it on the embedded @daytonaio sandbox.

Had it return an image artifact (which gets auto-ingested for later search).

This kind of analysis is amazing - and just within a min or two - using your preferred LLM, and fully customizable via prompting the agent.

[Kirk Marple @KirkMarple]

Another fun weekend project was adding @twelve_labs multimodal embeddings into @graphlit, and then integrating into @dossium.

Now you can search across any content via text, image and multimodal embeddings, and we cite where in the video the multimodal hits were found.

[Kirk Marple @KirkMarple]

https://t.co/MxKPITSkK1

[Kirk Marple @KirkMarple]

https://t.co/DEj3MT3dX0

[Kirk Marple @KirkMarple]

The next great “API company” in the agent era is the API to enterprise context.

Every product team wants agents/copilots that actually understand:
customer history, internal docs, Slack, email, CRM, meetings.

The hard part isn’t the chat UI.
It’s the connectors + ingestion + retrieval substrate underneath.

That’s what we built at @graphlit.

Now opening OEM + private deployments.

[Kirk Marple @KirkMarple]

We're excited to announce Dossium.

Your CRM knows who your customers are. @dossium knows everything about them - from every email, call, and Slack thread.

You have a call in an hour. You need to know what happened in the last meeting, whether there are open issues, and what their VP said in that email thread last month. So you open five tabs and start searching.

[Kirk Marple @KirkMarple]

We have been getting a great response from our Dossium announcement.

Customer interest from all kinds of verticals.

All asking to try out the beta.

We are working hard to get this out in early February!

If you’d like to guide the initial release with your requirements, please reach out via DM or setup a Zoom. (Link in article)

[Kirk Marple @KirkMarple]

https://t.co/upOsCRnkB4

[Kirk Marple @KirkMarple]

https://t.co/mr7jJmRwPm

[Kirk Marple @KirkMarple]

https://t.co/cJQvbDtC8j

[Kirk Marple @KirkMarple]

https://t.co/weN8SvK12t

[Kirk Marple @KirkMarple]

🆕 The power of @graphlit, with the beautiful UI of @zine_ai...

We are excited to launch Graphlit Studio this week.

30+ data sources, content and entity management, MCP-native chat, and multi-project support.

Showing our new visual workflow builder using @xyflowdev.

[Kirk Marple @KirkMarple]

🔥 Just tested on-demand entity enrichment with @p0 task groups API. The results are wild.

Enriched the same org (OpenAI) across 3 processor tiers (async). Started with simple JSON-LD of an Organization with name assigned.

See the JSON-LD screenshots in replies.

Base (60s):
• 15 employees with titles
• Ownership breakdown (MS 27%, Foundation 26%)
• 7 social accounts
• 9 source citations

Core (6-7 min):
• Everything in Base +
• Full street addresses
• NAICS codes
• More founders (7 vs 4)
• 11 citations

Pro (11.5 min):
• Everything in Core +
• 11 founders with personal URLs
• 6 international subsidiaries
• Corporate structure tree
• Multiple contact emails
• 13+ social accounts

🤯The surprise: Base async delivers 80% of Core's value in 1/6th the time. For knowledge graphs at scale, that's a game changer.

[Kirk Marple @KirkMarple]

@robinebers It’s just a commodity RAG service. Has tons of limitations which any other RAG API supercedes.

For the simple stuff, it’s great if you want to stay in the Google ecosystem.

But for real production content, you need a platform like @graphlit or one of the other vendors.

[Kirk Marple @KirkMarple]

Memory Is the Interface.
Knowledge Is the Substance.

🧵1/n

[Kirk Marple @KirkMarple]

@ExaAILabs And @graphlit: https://t.co/owL4WeZvEL

[Kirk Marple @KirkMarple]

If you hadn’t seen, we recently refreshed our brand websites as well.

@graphlit Semantic memory platform:

https://t.co/wCruGjOQmq

And our @zine_ai MCP client app:

https://t.co/5sJj0j7PP6

[Kirk Marple @KirkMarple]

It’s been interesting to see so many companies in the agents and memory space optimizing for onboarding.

“In five lines of code”
“Integrate in 5 min”

While that’s important to avoid friction, it’s not a concern we hear a lot at @graphlit

The questions we get are more real-world:
- can you dedupe content?
- how do I get a diff when content changes?
- can you support data source “XYZ”?
- how can I search for similar content with a time range?

If you need to integrate memory into your AI agents, let us know how @graphlit fits your needs. And if it doesn’t.