Earlier this month, Andrej Karpathy posted about how he now spends more LLM tokens building persistent knowledge wikis than writing code — markdown files incrementally compiled into a wiki by AI agents. It went massively viral. We've been building exactly this at Inkeep. And it's almost ready. Later this week, we're open-sourcing Open Knowledge — an agent-native knowledge platform where humans and AI co-create knowledge in real time. Think: Obsidian, but built for AI co-creation. → Markdown files in git — no database, no proprietary lock-in → Rich WYSIWYG editor with source toggle → Built-in MCP server — Claude Code, Cursor, Windsurf, Codex work out of the box → Real-time CRDT collaboration — AI agents and humans editing side by side → Two commands to get started Open source. MIT licensed. Launching on Product Hunt later this week.
Inkeep
Software Development
San Francisco, California 3,444 followers
AI Agents for Customer Operations.
About us
Inkeep helps companies ship AI Agents for customer experience, GTM, and operations teams. Inkeep Agents are deployed at scale today with leading companies like Anthropic, Midjourney, Clay, and PostHog to help with everything from customer support to in-product copilots to sales. Our Agent Engineering team works closely with Enterprises and high-growth teams to deliver on reliable agents quickly, but we also offer a full open source platform so your teams can own and manage agents in the long run. Accessibility is our goal: agents can be edited and managed by both engineers and business teams, all in a single platform. Business teams get an intuitive drag-and-drop visual builder while engineers get developer SDK with type safety, CI/CD, version control, and all the code-based tools they expect. Platform highlights: - A No-Code Agent Builder and Developer Framework with full 2-way sync - Unified search and RAG for knowledge bases, docs, and company data - MCPs and integrations with your APIs and software - Intelligent insights and monitoring over how Agents perform Backed by Khosla Ventures, Y-Combinator, Great Point Ventures, and other leading investors.
- Website
-
https://go.inkeep.com/INK-LN-PROF
External link for Inkeep
- Industry
- Software Development
- Company size
- 11-50 employees
- Headquarters
- San Francisco, California
- Type
- Privately Held
- Founded
- 2023
- Specialties
- generativeai, llms, search, neuralsearch, rag, support, retrieval, ai, ml, devtools, chatbot, genai, aichat, productmanagement, analytics, deflection, devex, customer engagement, self-service, customer support, customer service, customer engagement, customer success, technical support, support copilot, cx, customer engagement, customer experience, customer support, aiteammates, aiagents, agents, agent workforce, ai agents, ai assistants, ai teammates, unified search, enterprise search, workflows, agent workflows, business automation, multi-agent, no-code, agent builder, agent engineering, ai agents, gtm agents, sales agents, marketing agents, in-product agents, in-product copilots, operations, automations, low-code, agent framework, workflow builder, visual builder, conversational agents, chat assistant, and ai assistant
Locations
-
Primary
Get directions
156 2nd St
San Francisco, California 94105, US
-
Get directions
New York, NY 10003, US
-
Get directions
156 2nd St
San Francisco, California 94105, US
Employees at Inkeep
Updates
-
Missed our latest webinar with Composio? We walked through how to give AI agents access to 15,000+ tools using Inkeep + Composio MCP servers — without building integrations from scratch. ⚡ Live demos connecting agents to Slack, GitHub, and Gmail 🧠 Best practices for production-ready agent integrations Watch the recording 👇 https://lnkd.in/g8KSiXd2
Empowering AI Agents with 15K+ Tools using Inkeep and Composio
https://www.youtube.com/
-
Inkeep reposted this
We signed our first real office in 4 hours. 7 months ago, we at Inkeep were in a 'co-working' space: But the reality was: - Sales teams whispering on prospect calls. - 2 Zoom calls happening two feet apart. - Constant anxiety of eavesdroppers listening to our calls We’d outgrown our co-working with startups. But we weren’t ready for a multi-year lease either. What we needed was weirdly specific: • Flexible terms • Quiet rooms to actually think • Transit access we’d use • And yes… an ocean view to survive early-stage chaos We started talking to commercial brokers. We told JLL straight up: “We’re not ready for a big commercial lease.” This is where most firms would lose interest. Instead, they leaned in. They’ve built deep relationships with WeWork and other coworking operators. Their thesis is simple: Help the startup now, then earn the relationship later. Will Cassriel from JLL packed a full day of tours into one afternoon. Jonathan Smith (my Chief of Staff) flew out, and we walked every serious option. By sunset, we had our pick. But the real value wasn’t the tour. - They negotiated on our behalf. - Explained contract language we’d never seen. - And recently helped us extend our WeWork agreement so we could buy more time before making a bigger bet. For any founder navigating office decisions, I'd recommend choosing partners who optimize for the relationship, not the transaction. You can feel the difference immediately.
-
-
🚨 Happening this Thursday If you’re building AI agents that need to connect to real-world tools- this is for you. This Thursday, we’re going live with Inkeep + Composio to show how you can give AI agents access to 10,000+ tools without maintaining brittle, one-off connections. 🗓 Feb 26 ⏰ 10:00 AM PST 💬 RSVP in the comments What you’ll learn: • How Inkeep’s AI agent framework handles tool use and agentic workflows • Why integrations are critical for production-grade AI agents • How Composio’s MCP servers unlock 10,000+ tools out of the box • Live demo: connecting agents to Slack, GitHub, and Gmail • Best practices for auth, testing, and deploying agent integrations in production Speakers: 🎤 Omar Gonzalez — Founding Engineer, Inkeep 🎤 Jayesh Sharma — AI Engineer, Composio 🎤 Gaurav Varma — DevEx Engineer, Inkeep (Moderator) If you're serious about shipping real-world agents- don’t miss this. 👇 RSVP in the comments
-
-
🌟 Excited to see Descope's Docs MCP Server live! AI-powered IDEs are quickly becoming the default way developers build. Making product knowledge directly accessible inside those workflows is a big step forward, which is why we built an MCP of our own to help you 'vibe code' agents using our agents SDK (link in the comments). We’re proud that Inkeep is powering key under-the-hood components of Descope's MCP server with semantic search and RAG. Congrats to the Descope team on the launch 👏
🚀 Introducing the Descope Docs MCP Server If you’re using AI-powered IDEs (who isn’t?), you now have an easy way to add auth to your apps and reference Descope product knowledge. Our Docs MCP Server is a hosted MCP server that gives AI agents and MCP-compatible tools direct, structured access to Descope’s product knowledge right where you already work. ✅ Troubleshoot errors and misconfigurations ✅ Answer architectural and design questions ✅ Get direct docs excerpts and supporting references ✅ Tell your AI coding assistant to add Descope auth, widgets, authorization, and more to your app Links to get started are in the comments 👇 P.S. Thank you to Inkeep for powering key under-the-hood components of this MCP server with semantic search and RAG!
Descope Docs MCP Server
-
We think "Agent Engineer" is becoming a real role. After hundreds of hours building AI agents, we've condensed what we've learned into a new guide: • How to structure prompts that actually work • Troubleshooting when agents fail • Coordinating multiple specialists Not theory. Battle-tested techniques. Link in the comments below.
-
-
Inkeep reposted this
Excited to be presenting tomorrow at Auth0's CampAI night! I'll be sharing a live demo and keynote on how teams can build Agents that use MCP tools on the Inkeep platform. Thank you to Shreya Gupta and Eli Rabek for organizing this. Link to sign up in the comments 👇️ 👇️ 👇️
-
-
Inkeep reposted this
Want a sense of what Heads of AI are thinking today? Consolidation. A pattern we see over and over is that when AI assistants first rolled out around the GPT-3.5 and GPT-4 wave, teams saw quick wins that felt transformational at the time. Those wins were real, but they were narrow, tied to specific domains or workflows, and rarely designed to generalize beyond the exact problem they were bought for. As teams tried to extend those wins to adjacent functions, the answer was almost always another tool, another vendor, another contract. Over time, this created a patchwork where context lived in silos, AI could not reason end-to-end, and humans quietly became the glue doing the real work by stitching systems together manually. The irony is that the more “AI” teams adopted, the more operational drag they introduced, and the less leverage they actually gained. Fast forward two years and many Heads of AI are now managing bloated stacks with upcoming renewals, uneven technical depth, and very little real autonomy in the system. When VPs come to agentic platforms like Inkeep, what actually drives adoption is not novelty, but the ability to connect tools, reason across tasks, and take action in real workflows. That is why I believe we are approaching an inflection point, not just in AI adoption, but in whether AI finally starts doing meaningful work instead of creating more of it.
-
Inkeep reposted this
Fingerprint increased user activation 18% and cut support tickets >58% in under 6 months without adding a single support hire. They did it by making their existing documentation actually work for users. Here's how they pulled it off with Inkeep. 👇 When we first connected with Alvin, Fingerprint Director of Growth, they were facing a familiar scaling wall.Their docs were comprehensive. Users even cited documentation quality as a reason for converting to paid plans. But there was a problem: discoverability was broken. "We had a very comprehensive set of docs, but they weren't the most easy to explore," Alvin explained. "We weren't sure how quickly people were getting answers and if they were even finding their answers." They needed to scale support without scaling headcount. That meant making self-serve actually work. What changed: (1) Deployed AI-powered search that could parse their entire doc library and surface precise answers in seconds (2) Built a feedback loop showing exactly where documentation had gaps (3) Integrated across docs, help desk, and product for a unified experience The results went beyond support deflection: >48% reduction in tickets (validated via A/B test). 18% lift in activation rates measured by first API calls almost immediately. Monthly doc improvement cycles running across 3+ teams. As Alvin put it: "Inkeep provides feedback loops to improve our docs and identifying the gaps was probably the biggest thing we were able to solve." The best part is that results have only improved since they've rolled out our Zendesk co-pilot. Link in the comments to read the full Fingerprint case study.
-
-
🚀 Webinar: Connecting AI agents to 15,000+ real-world tools- without painful integrations with Inkeep and Composio AI agents are only as powerful as the tools they can use. But connecting agents to CRMs, productivity tools, databases, and APIs often means maintaining dozens (or hundreds) of brittle integrations. In this webinar, engineers from Inkeep and Composio will show how you can give AI agents instant access to 15,000+ tools using Composio’s MCP server library- without building integrations from scratch. 🗓 Feb 26 ⏰ 10:00 AM PST 💬 RSVP in the comments What you’ll learn: • How Inkeep’s AI agent framework handles tool use and agentic workflows • Why integrations are critical for production-grade AI agents • How Composio’s MCP servers unlock 15,000+ tools out of the box • Live demo: connecting agents to Slack, GitHub, and Gmail • Best practices for auth, testing, and deploying agent integrations in production Speakers: 🎤 Omar Gonzalez — Founding Engineer, Inkeep 🎤 Jayesh Sharma — AI Engineer, Composio 🎤 Gaurav Varma — DevEx Engineer, Inkeep (Moderator) If you’re building real-world AI agents (or want to), this one’s for you. 👇 RSVP in the comments
