OpenCrab
OpenCrab is an ontology-first Graph RAG platform that turns documents into reusable, traceable knowledge packs
OpenCrab's bet is that knowledge you plan to package, license, or rerun as an agent workflow deserves structure, not just an embedding. The nine-space MetaOntology, versioned packs, and a personal MCP URL that Claude, ChatGPT, and Cursor can all share are a genuinely different shape from ordinary RAG, and at $10/mo for private ingest the experiment is cheap enough to settle the question yourself. If you just want to ask questions of a pile of PDFs, the up-front modeling is overhead you will resent — reach for a lighter vector store instead.
Verified 11d ago · liveness 74/100 · cite: rightaichoice.com/tools/opencrab
- Researchers and analysts building a private evidence graph from papers and reports
- Domain experts packaging expertise into ontology packs they can publish and license
- AI builders grounding Claude, ChatGPT, or custom agents with MCP-connected knowledge
- Teams that need answers traceable to specific source evidence
- Buyers who want a turnkey vector database and no ontology modeling
- Non-technical users expecting zero-configuration chat over their files
- Single-user one-off document Q&A where a lighter RAG tool would suffice
We scan live Reddit threads, YouTube comments, X posts, G2 reviews and other communities — and hand you an honest verdict in under a minute.
- Honest verdict, not marketing
- Real pros & cons from real users
- Attributed quotes with receipts
3 free scans · no card needed
Skip OpenCrab if you want to ask questions of a folder of PDFs and never think about entities, edges, or evidence modeling again — the ontology is the product here, so a lighter RAG tool will get you to an answer sooner.
Private ingest, which is the whole point for most buyers, does not exist on Free — any of your own documents, URLs, or data ZIPs require the $10/mo Pro tier.
Pro at $10/mo sits under most structured-knowledge tools and is cheap enough that a solo researcher can test private ingest for a month. Expert at $30/mo is where a working pack publisher lives; it is priced like a light SaaS seat rather than an enterprise knowledge platform. Below OpenCrab, a plain vector store is cheaper but gives you no ontology, and above it, graph and knowledge-graph platforms charge per seat or per volume. Company-level permissions and multiple seats only come via
In short
OpenCrab — OpenCrab is an ontology-first Graph RAG platform that turns documents into reusable, traceable knowledge packs. Best for Researchers and analysts building a private evidence graph from papers and reports, Domain experts packaging expertise into ontology packs they can publish and license, AI builders grounding Claude, ChatGPT, or custom agents with MCP-connected knowledge. Free to start; paid plans from $10/mo.
What's new in OpenCrab
Checked 3 days agoAcross the latest 1 update: 1 changelog entry.
What people actually say about OpenCrab — is it worth it?
We ran a structured research pass across product reviews, community discussions, and post-purchase forum threads to surface the patterns vendors won't publish themselves. Below: the recurring strengths, the hidden costs people mention most, and the cohort that consistently regrets adopting this tool.
39 mentions across 4 sources (Hacker News, YouTube, Product Hunt, GitHub) · researched Aug 27, 2026.
Average across the 4 sources that answered — each source counts once, not each post.
- +Provides auditable, evidence-grounded answers by tracing every response to source material.
- +Ontology-first design transforms knowledge into reusable, sellable packs.
- +MCP integration enables seamless use with Claude Desktop, GPT, and custom agents.
- +Supports diverse document ingestion including PDF, HWP, GitHub, and expert notes.
- +OCR and CLIP handle scanned PDFs and images effectively.
- −Ontology modeling adds upfront complexity and a learning curve.
- −Cron jobs execute in wrong profile environment, causing misconfiguration.
- −Terminal launch crashes since v0.2.66 due to CUDA assertion.
- −Background-task results are not visible in chat, reducing transparency.
- −Community feedback is extremely limited outside GitHub.
- • Potential cost for mobile apps or advanced MCP features not in free tier.
- • Pricing unclear; teams may need to contact sales.
Viability Score
How well maintained and how widely used is OpenCrab? Built from what the vendor actually publishes (docs, changelog, tutorials, integrations, pricing), whether the site is live, and how much real users discuss it. How we calculate this
Last calculated: October 2026
How we score →Key Features
- Ingest documents from PDF, URL, Notion, GitHub, and expert notes
- Automatic parsing, OCR, and chunking on ingest
- Nine-space MetaOntology extraction across Entity, Evidence, Claim, Policy, and Lever
- Graph RAG querying with answers traced back to source evidence
- Versioned ontology packs built from graph nodes, edges, and evidence
- Pack sharing with teams and publishing to the marketplace
- Personal MCP URL for external AI clients
- Connect the same MCP URL to Claude, ChatGPT, Cursor, Codex, Claude Code, and Antigravity
- Reverse ingest from connected MCP tools back into a pack
- Crab Agent for agent-driven workflow runs
- Pack QA and project-style workflows on Expert
- Qdrant output connector for downstream vector retrieval
- Runtime panels across webapp, desktop app, and MCP
- Deployment to Codex and Claude Code
- Bilingual Korean and English interface
About OpenCrab
OpenCrab is an ontology-first Graph RAG platform from AgentKorea for teams that want knowledge they can package, audit, and reuse rather than just retrieve. Instead of dropping files into a vector index, it ingests PDF, URL, OCR, Notion, GitHub, and expert notes, then structures the contents into nodes, edges, claims, and evidence that answers can be traced back to. Work runs through four public stages — ingest, structuring, saving the result as a versioned pack, then querying — with parsing, OCR, and chunking handled automatically at the front end. Structuring uses a nine-space MetaOntology taxonomy that names Entity, Evidence, Claim, Policy, and Lever among its categories. You query the finished graph through Graph RAG inside OpenCrab or over a personal MCP URL that plugs the same graph into Claude, ChatGPT, Cursor, Codex, Claude Code, and Antigravity. A Crab Agent, pack QA, and project-style workflows sit on the Expert tier, and reverse ingest pulls results from connected MCP clients back into an OpenCrab pack. The interface is bilingual Korean and English. Pricing is self-serve through Paddle and tiered by what you need to do, not by seat count: Free at $0/mo for exploring and querying public ontology packs and buying expert-made paid packs, Pro at $10/mo for private document, URL, and data-ZIP ingest with an isolated personal graph RAG workspace, and Expert at $30/mo for pack publishing, QA, Crab Agent, and advanced reverse ingest. The guide also lists Enterprise for company-level permissions and multiple seats via inquiry. Against lighter RAG stacks that index and retrieve with little structure, OpenCrab asks for more modeling work up front because the ontology is the product. That trade favors researchers, domain experts, and AI builders who intend to sell, license, or repeatedly run against the knowledge they assemble, and it gets in the way of anyone who wants a zero-configuration store for chat.
Behind the Verdict
What OpenCrab actually sells is the pack. You ingest documents, URLs, notes, or code; the platform parses, OCRs, and chunks them; then a nine-space MetaOntology pass extracts entities, concepts, evidence, relationships, and claims and wires them into a graph where every answer carries its source. Once that graph exists, it is saved as a versioned pack you can share with a team or list on the marketplace. That is the part that differs from an index-and-retrieve stack: the graph is portable, auditable, and reusable, so the same knowledge can serve a research memo today and an agent's long-term memory next quarter. The MCP layer is the practical payoff. You generate a personal URL under /api/mcp/ and register that one endpoint in ChatGPT (Settings > Apps), Claude (Customize > Connectors), Codex (~/.codex/config.toml), Claude Code (claude mcp add --transport http), and Antigravity. Permissions follow your current account tier rather than being frozen at URL creation, so upgrading to Expert can unlock Expert-level tools through URLs you already handed out. Two cautions the docs themselves raise: the access key lives inside the URL, so treat it like a secret and Disable and regenerate if it leaks, and a new URL means re-pasting it into every connected tool. Where it fits: researchers building a private evidence graph they can verify against sources, domain experts turning a specialty into a pack they can sell or license through the marketplace, and AI builders who need a grounded knowledge endpoint behind an agent. The Crab Agent, pack QA, and project-style workflows on Expert point at people running this as a small publishing or delivery operation, and the quarterly Korean-language origin shows in the bilingual interface. Where it does not: if you want a zero-configuration chat over a folder of files, the ontology modeling is work you will not recover. Single-user one-off document Q&A is cheaper on a lighter RAG tool, and teams that need a broad public REST API surface for custom service integration should check scope before committing. Selling packs is gated behind Expert, an OpenCrab contract and review, and a seller profile — budget for that review step rather than assuming instant listing. Note also that the platform is small-team (AgentKorea) and support runs through email and community channels, which is fine for a $10/mo experiment and a real consideration for anyone standardizing a team on it.
Researching OpenCrab? Get your full AI stack in 60 seconds.
Free, no signup — tell us your goal and get tools matched to your budget & existing stack.
Real-world workflow fit
Concrete scenarios for the personas OpenCrab actually fits — and what changes day-one when you adopt it.
Sign up, land on the dashboard, install a public pack from the Marketplace to learn the format, then upgrade to Pro and ingest a folder of PDFs and report URLs so OCR and chunking run automatically into a private graph.
Outcome: A personal evidence graph you can query where every answer points back to the source paragraph, instead of a folder of PDFs you have to re-read.
Go to Expert, ingest your own notes and reference material, structure it with the nine-space MetaOntology, run pack QA, then publish through the marketplace contract and review flow.
Outcome: A listed, versioned ontology pack buyers can install and query, with OpenCrab handling the sales and delivery path.
Open the MCP page, generate a personal URL under /api/mcp/, register it in Claude (Customize > Connectors) and Codex (~/.codex/config.toml), and point an agent at the graph for retrieval.
Outcome: One shared knowledge endpoint behind several clients, with retrieval grounded in your graph rather than the model's training data.
Use Cases
- Ingest papers and reports into a private evidence graph where each answer cites its source
- Package domain expertise into a versioned ontology pack and list it on the marketplace
- Register one OpenCrab MCP URL in Claude, ChatGPT, and Cursor so all three query the same graph
- Build a content planning or brand analysis ontology instead of re-reading scattered briefs
- Ground an agent's long-term memory on a structured graph rather than a flat vector index
- Push the finished graph to Qdrant for downstream vector retrieval
- Save Codex or Claude Code project results back into a pack through reverse ingest
- Query public ontology packs for free before committing to private ingest
Models Under the Hood
as of 2026-10-08
Limitations
- Free is a viewing and buying tier, not an ingest tier: private document and text ingest opens at Pro, and in the guide a blocked Ingest page on Free is described as expected behavior.
- Selling packs, running pack QA and Crab Agent, and advanced reverse ingest are Expert-only, and pack publishing also needs Expert permissions, a seller profile, and OpenCrab's contract and review step, so plan for that lead time rather than listing same-day.
- The docs recommend keeping each MCP URL private and using Disable plus regeneration if it leaks, which means re-registering the new URL in every connected tool.
- OpenCrab is built by a small team (AgentKorea) with support through email and community channels, and external AI clients have their own menu names and permission rules that change between their releases.
as of 2026-09-27
Verification history
We have re-verified OpenCrab 8 times since . Each pass re-reads the vendor's own pages and re-checks every listed field against that evidence; passes where nothing had changed are marked as such.
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
Showing the 6 most recent of 8 verification passes.
Free to cite with attribution — this page re-verifies continuously.
12-month cost
Project the real annual outlay, including the implied monthly cost when only an annual tier is published.
Vendor list price only. Add-on usage, seat overages, and contract minimums are surfaced under Hidden costs & gotchas.
Plans compared
For each published OpenCrab tier: who it actually fits, and what it adds vs. the previous tier. Cross-reference the cost calculator above for projected annual outlay.
Free
$0/mo
Ideal for
Someone who wants to browse and query public ontology packs, or buy expert-made paid packs, before spending anything
What this tier adds
Starting tier at $0/mo: public pack access, basic graph search, paid pack purchase, and lookup from external AI tools — private ingest and personal pack creation are locked.
Pro
$10/mo
Ideal for
A researcher or analyst who needs their own PDFs, URLs, and data ZIPs turned into a private, isolated knowledge pack
What this tier adds
Adds private document, URL, and data ZIP ingest plus a personal graph RAG workspace with isolated storage, and lets you connect your own packs to external AI tools over MCP.
Expert
$30/mo
Ideal for
A domain expert or small operation publishing, QA-ing, and selling ontology packs, or running agent workflows against them
What this tier adds
Adds pack publishing and selling after OpenCrab contract and review, pack QA, Crab Agent, project-style workflows, and advanced reverse ingest from connected MCP tools.
Where the pricing makes sense
The company stage and team size where OpenCrab's pricing actually pencils out — and where peers do it cheaper.
Pro at $10/mo sits under most structured-knowledge tools and is cheap enough that a solo researcher can test private ingest for a month. Expert at $30/mo is where a working pack publisher lives; it is priced like a light SaaS seat rather than an enterprise knowledge platform. Below OpenCrab, a plain vector store is cheaper but gives you no ontology, and above it, graph and knowledge-graph platforms charge per seat or per volume. Company-level permissions and multiple seats only come via
Setup time & first value
How long it actually takes to get something useful out of OpenCrab — broken out by persona, not the marketing-page minute.
Roughly 3 minutes to create an account and reach the dashboard, 5 minutes to read the tier matrix and pick Free, Pro, or Expert, and about 5 minutes to install a marketplace pack. Generating an MCP URL takes another 3 minutes, and registering that URL across ChatGPT, Claude, Codex, Claude Code, and Antigravity is budgeted at around 10 minutes — though plan for extra time on Team or Enterprise
Switching to or from OpenCrab
How to bring data in from common predecessors and how to get it back out — written for the switcher, not the buyer.
- →From a plain vector store: keep the store for fast retrieval and re-ingest the same documents through OpenCrab so entities, relations, and evidence are extracted on top of the chunks.
- →From folder-and-search document workflows: upload the PDFs, URLs, and notes to Pro, let parsing, OCR, and chunking run, and start querying the resulting graph.
- →From chat-over-PDF tools: move the source documents into a Pro ontology pack when you need answers tied to a specific cited source rather than a fluent paragraph.
- →From an MCP-connected client: generate an OpenCrab MCP URL and register it alongside your existing tools so the same graph is reachable from each client.
- ↗To a plain vector database: use the Qdrant output connector to push the finished graph out for retrieval, keeping the vector store as the serving layer.
- ↗To a lighter RAG tool: export your source documents and accept the loss of nodes, edges, and evidence traceability if you only need fast Q&A.
- ↗To direct model APIs: skip the graph layer and send documents straight to the model when traceability to a specific source is not a requirement.
Integrations
Resources & Guides
Tutorials & Learning
YouTube returned 6 videos for “OpenCrab”, and we withheld 6: 6 could not be judged, because “OpenCrab” is a single word that other videos use for other things. We are showing none, because we could not prove any of them are about OpenCrab.
Official links
Tools that pair well with OpenCrab
Common stack mates teams adopt alongside OpenCrab, with the specific reason each pairing earns its keep.
GraphRAG
Open-source Microsoft Research RAG that builds a knowledge graph and community summaries from your documents, then queries them in four modes.
Iris.ai
Auditable AI knowledge layer for regulated enterprises that must defend every answer in an audit.
Klippa
Doxis (ex-Klippa) turns invoices, receipts, and IDs into structured data with OCR, verification, and spend control.
Featured Head-to-Head Comparisons
Opencrab vs Praktika
Praktika and OpenCrab serve completely different needs. Choose Praktika for conversational language practice with AI tutors; choose OpenCrab for structuring messy knowledge into ontology packs for AI agents. They are not direct competitors.
Opencrab vs Screenplayiq
If you're a screenwriter or studio exec needing script marketability analysis and box office predictions, ScreenplayIQ is the clear choice with its beat sheets, heatmaps, and comparative market data. But if you're a researcher or AI builder who needs to turn messy documents into structured, auditable knowledge graphs with MCP integration (Claude/GPT), OpenCrab is far more powerful. Choose based on your domain: narrative finance vs. ontology-driven knowledge packaging.
Alternatives to OpenCrab
View all
Frequently Asked Questions
Categories
Used OpenCrab? Help shape our editorial sentiment research.