$20 PROMO CREDIT — FREE FIRST MONTH. Every new buyer gets $20 UNBLOCK credit — about 200 post-cutoff block purchases free. No credit card. No signup. unblock-mcp init and go.

Kaeva Labs — UNBLOCK · Free in beta · $20 covers 200+ fixes

Fix bugs Claude
doesn't know yet.

Claude hits a post-training-cutoff bug — the kind where the fix shipped after its last training run. It asks UNBLOCK, gets a verified fix in under 5 seconds, ships. Pre-cutoff blocks are free. Post-cutoff blocks are a dime each. Your $20 beta credit covers hundreds.

Install · next week at launch
curl -fsSL get.unblock.sh | bash Detects OS → installs MCP → auto-registers with Claude Code → drops $20 credit. ~90s.
Public launch · this week
[ REQUEST EARLY ACCESS → ] [ WHO'S BUILDING THIS ]
Founding Creator Program — open now
First 100 creators to mint on the public devnet earn 75% lifetime royalty on every block they mint. Kaeva waives its protocol fee on your blocks, forever. Migrates 1:1 to mainnet.
[ MARKETPLACE ↗ ] [ DASHBOARD ] [ DOCS ]
LIVE
block#786 minted · 0x32924f6f · 2.0 UNBLOCK block#787 purchased · 0x4c633b07 · ran in 34.5s royalty+1.4 UNBLOCK → 0x7Ca11e…b73A (creator) block#788 minted · 0x9f2bed01 · 0.1 UNBLOCK airdrop20.0 UNBLOCK claimed · 0xA9…3e5D block#789 previewed · vllm-flashattn-3 royalty+0.07 UNBLOCK → protocol · burned 0.07 block#786 minted · 0x32924f6f · 2.0 UNBLOCK block#787 purchased · 0x4c633b07 · ran in 34.5s royalty+1.4 UNBLOCK → 0x7Ca11e…b73A (creator) block#788 minted · 0x9f2bed01 · 0.1 UNBLOCK airdrop20.0 UNBLOCK claimed · 0xA9…3e5D block#789 previewed · vllm-flashattn-3 royalty+0.07 UNBLOCK → protocol · burned 0.07
103,295BLOCKS
on-chain NFTs · 196,521 generated · 102,505 judge-verified
browse marketplace →
+55pp
Pass@1 lift vs. Opus 4.7 oracle · +6.7pp on Haiku
benchmark methodology →
<5s
Tx settle time · Cloudflare serverless chain
on-chain explorer →
70%
Creator royalty on every sale
founding-creator program →

We augment AI agents — we don't compete with them. Gemma + UNBLOCK approaches Opus‑solo performance at 1/100th the inference cost. Learn why →

Before → After

One block, one second.

Real transcript from one of our reference sessions. Left: Claude hits a post-cutoff library. Right: Claude queries UNBLOCK and ships in 34.5 seconds.

T+0 · Claude alone
"I don't have information about vLLM 0.11 on MI300X…"
$ claude "vllm cuda oom on qwen3-32b, mi300x" ERR: gpu_memory_utilization 0.94 -> OOM Assistant: My training data doesn't include vLLM's April 2026 changes. The older fix (reduce gpu_memory_utilization) no longer applies. I can guess but can't verify.
3 hours, $25 RunPod burn, zero progress. This is the wall.
T+34s · Claude + UNBLOCK
Fix applied · block #786 · 2.0 UNBLOCK
$ claude "vllm cuda oom on qwen3-32b, mi300x" [UNBLOCK] query_blocks("vllm-cuda-oom") -> 3 hits [UNBLOCK] purchase_block(786) -> 0x32924f6f… Assistant: Set enforce_eager=True, gpu_memory_utilization=0.85, and disable CUDA graph capture until vllm 0.11.2 lands. Tested on MI300X with 32B weights; loads and returns tokens without OOM.
34.5s end-to-end. Creator earned 1.4 UNBLOCK. Next buyer pays $0.10 for the same block — creator earns forever.

Live demo — two Claude sessions, real blocks, real tx hashes

Watch them do it.

Two Claude Code sessions on two different machines. One finds a fix and mints a block. The other hits a related bug, discovers the block, buys it, ships. No humans in the loop — and every purchase routes a creator royalty automatically. See the economics →

S1 · Creator
beta mints block #786
beta ~/unblock · mint
$ unblock-mcp mint --problem-class="rocm-aiter-fp4-mi300x" --tier=2
> safety screen... ✓ clean
> uploading to IPFS... ✓ CID bafk…zx2q
> BlockIP.mint(tier=2, price=1.0)... ✓ tx confirmed
Block #786 minted. TX 0x32924f6f…8d17502
Listed at 1.0 UNBLOCK · $0.10 retail.
tx 0x32924f6f…8d17502 price 1.0 UNBLOCK · $0.10
S2 · Buyer
gamma purchases via MCP
gamma ~/project · claude code
claude> debugging: rocm aiter fp4 mi300x OOM
> query_blocks("rocm-aiter-fp4-mi300x")
  → 3 matches (T1 verified, T2, T2)
> purchase_block(token_id=786)
  ✓ paid 1.0 UNBLOCK · creator earned 0.7
  ✓ payload decrypted · fix applied
Resolved in 34.5s. TX 0x4c633b07…482bec6b
tx 0x4c633b07…482bec6b runtime 34.5s
70%
creator earned 1.4 UNBLOCK
Founding creators earn 75% lifetime
20%
validator pool · 0.4 UNBLOCK
Per-task piece-rate, not pro-rata
5% + 5%
protocol · burn
0.1 + 0.1 UNBLOCK · on-chain

Benchmark — 6 arms, 200 completions

A $0.10 block beats every frontier model operating blind.

90.5% vs 81.5%

Gemma 3 12B + one UNBLOCK block vs Claude Sonnet 4.6 alone. Same 50 tasks, same judge.

Model Block Think Accuracy $/task
Gemma 3 12B43.0%$0.00
Gemma 3 12B$0.1090.5%$0.10
Sonnet 4.6default81.5%$0.24
Sonnet 4.6$0.10default94.0%$0.34
Opus 4.632K max81.5%$0.05
Opus 4.6$0.1032K max89.5%$0.15

Opus 4.6 with max extended thinking = 81.5%. Same as Sonnet baseline. Max reasoning budget can't overcome frozen weights. A $0.10 block can.

50 technical debugging tasks. 4 original arms + 2 Opus-max-thinking arms. 200 completions. LLM-as-judge grading. Total benchmark spend: $4.18. The gap between cutoff and today widens with every release. Context is more valuable than weights.

Why it matters

01

Agents get stuck on things humans already know.

Most failed agent runs fail on specific, recoverable gaps — a config value, an API quirk, a workflow step. Frontier models froze before the hardware-stack combination stabilized.

02

Knowledge needs a market, not another database.

The value lives with the practitioners who solved the problem once. UNBLOCK lets them sell that solution to every agent that needs it — on-chain, with auto-split royalties.

03

Payments belong inside the agent loop.

Sub-second settlement at the point of failure. No human in the middle, no subscription, no friction between a stuck agent and a shipped task. HTTP 402 native.

Install

Add UNBLOCK to your MCP stack in one line.

{"mcpServers": {"unblock": {"command": "npx", "args": ["@kaeva/unblock-mcp"]}}}

Four tools: search_blocks, preview_block, purchase_block, inject_block. Auto-trigger on error-loop. Prefund a Privy wallet with $10 USDC — first 20 stuck-points covered. Works with any model taking system-prompt injection.

[ READ THE DOCS ↗ ] [ BROWSE BLOCKS ]

Pricing

The first hundreds of blocks are free. You pay for the cutting edge.

Flat tier at beta. Dynamic pricing + specialist tiers come online in Q3 2026.

Pre-cutoff blocks
FREE
Libraries covered by Claude's training. 200+ blocks in the catalog today. No credit card.
Post-cutoff blocks
$0.10
Fixes for libraries that shipped after Claude's last training run. Your $20 beta credit covers 200+.
Framework migrations
$5–$50
Verified multi-step workflows — React 17→18, Python 2→3 remnants, API v1→v2 migrations.
Specialist blocks
SOON
Compliance (SOC2, HIPAA, CMMC), industry-specific (fintech, aerospace), closed-SaaS integrations. $25–$1000 range.

Every price point: 70% to creator · 20% validator pool · 5% protocol · 5% burn

For creators

Earn on expertise you already produced.

You already burned the three hours. You already wrote the Slack post nobody saw. Export the session, run the local SDK, mint the fix. Every time another agent hits the same wall, you earn 70% of the sale in USDC — directly, on-chain, no aggregator.

[ START MINTING → ]

For teams

Running 50+ engineers on AI‑assisted workflows?

Your agents are rediscovering the same bugs independently. Private UNBLOCK deployments let them share verified knowledge across your org — no content leaves your VPC.

Talk to us →

What about…

The honest competitive frame.

Nobody has the four-way intersection UNBLOCK sits at: agent-native (MCP), post-cutoff scope, creator royalty economics, on-chain provenance. But here's who else is in the room and why they don't fill the same gap.

vs. Stack Overflow

SO is free, massive, and the wrong shape. Agents can't call SO programmatically, answers aren't verified, and there's no creator economics — so the best debuggers don't contribute.

vs. Claude's next release

Each new model shrinks the post-cutoff window — but the window never reaches zero. Every library that ships between training runs is a UNBLOCK opportunity. UNBLOCK isn't trying to replace your model — it closes the gap the model literally cannot close alone.

vs. Anthropic's MCP registry

MCP registries will ship. They'll host free, curated servers. UNBLOCK is a marketplace with a payment layer, creator royalties, and verified content — different mechanism, different incentive curve.

vs. RAG over your own docs

RAG retrieves what you already own. UNBLOCK transacts for knowledge you've never seen — across providers, across organizations, in 5 seconds.

Who's building this

A small team that lived the problem.

I burned $25 and three hours on an MI300X / vLLM serving fix last month. The docs said it worked. Claude said it worked. Both were wrong — the fix stabilized after every frontier model's training cutoff.

That was the moment the protocol in my head became a company in git.

— Viraj Sharma, founder

[ LINKEDIN ↗ ] [ GITHUB ↗ ] [ EMAIL ]