Skip to main content

​
Become a Data Broker

What $500/year subscription could YOUR tool replace for $0.10/response? What fragmented API mess could YOUR tool normalize for $0.001/call? CoinGlass. Glassnode. Hyblock. BamSEC. These prosumer subscriptions gatekeep one or two premium features that power users actually pay for every month. You can unbundle any single one of those features, as curated intelligence for the Context app, as normalized data for SDK developers, or both. Context Protocol is a decentralized marketplace where AI agents pay for data. Build an MCP tool that delivers value, and you earn 90% of usage fees (Query responses and eligible Execute calls). The long-term vision: thousands of unbundled feature slices, shipped by independent contributors worldwide, collectively rival Bloomberg, FactSet, and Capital IQ as a decentralized, user-generated intelligence layer for AI agents. You own the piece you ship.
What we fund: feature-level MCP products that unbundle one paid workflow into either:
  • Curated intelligence (Query): a premium answer people would pay for on demand
  • Normalized infrastructure (Execute): a narrow typed primitive developers would rather rent than build
What we don’t fund: broad wrappers, dead categories, generic public API proxies, or products with no clear buyer. If you cannot point to a paid workflow people already spend money on, or to a painful integration developers already complain about, do not apply.

​
The Grant Pool

We’re allocating $10,000 USD to bootstrap the supply side of the Context Marketplace.

​
Funding Tiers

​
Tier S: $1,000 + Revenue Share

High-value tools that replace expensive subscriptions or painful integrations.
Tier S marketing support (conditional). If we judge your tool to be genuinely high-value, the kind we would personally pay for and recommend to a buyer, we will feature it in our demand-side demo library and help market it to our first cohort of paying users. This is not automatic for every Tier S grant. Most funded tools to date do not clear this bar. If you want yours to, pick a premium slice that unambiguously beats free chat, ship it cleanly, and we will do the rest.

​
Tier A: $500 + Revenue Share

High-utility tools with clear buyer value and good reliability.

​
Tier B: Revenue Share Only

Simple wrappers and adapters. No upfront grant, but immediate listing. Start earning from day one.
All tiers earn revenue share. Even Tier B tools generate USDC the moment agents use them. You keep 90% of every fee.

​
Request for Builders (RFB)

We no longer fund broad category wrappers like “crypto analytics,” “competitive intelligence,” or “financial research.” We fund feature-level products with:
  • a clear buyer
  • a painful workflow
  • an expensive current substitute
  • a narrow, testable surface contract
  • an obvious reason to exist on Context instead of as a generic dashboard or API wrapper
The best RFBs are not organized by protocol mode first. They are organized by buyer pain and workflow. Mode still matters, but it should usually be a property of the product:
  • Query: the buyer wants a premium answer with evidence
  • Execute: the buyer wants a narrow typed primitive
  • Both: the same underlying data can support both
The north star: submit a premium feature, not a generic category.Good: BTC/ETH exchange-flow regime intelligenceWeak: Crypto analyticsGood: Cross-venue perp funding / OI / liquidation normalizerWeak: Exchange API wrapper

​
Strong RFBs Include

  • Buyer
  • Painful workflow
  • Paid current substitute
  • Mode fit: Query, Execute, or both
  • 3-5 must-win prompts or methods
  • Evidence or schema contract
  • Tight scope boundary
What weak applications still get wrong
  • broad claims like “Bloomberg for AI” or “CoinGlass for agents”
  • no real buyer, just “this data is interesting”
  • no wedge versus free search, dashboards, or manual workflows
  • too many sources and too much normalization for a solo builder
  • products that are really alerting feeds or dashboard SaaS products
  • raw data wrappers with no evidence layer, no schema discipline, and no differentiation

​
Current Priority Areas

Right now, we are most excited about premium crypto and financial-data workflows because users in those markets already pay for better information and can justify paying per answer or per call.
Before you apply: the agent-native sanity check.A strong RFB unbundles a slice of a premium subscription into an answer an agent would naturally be asked. Before you build, write the exact sentence a buyer would type, and name the audience that would pay for the answer.The slice works if at least one of these audiences is clear:
  • a prosumer or independent buyer with no desk access (CoinGlass / Glassnode / BamSEC / FinTwit readers, small operators, indie investors)
  • a desk analyst who would still prefer typing one question into an agent over navigating a terminal menu for this specific query, even while keeping their terminal
  • an AI agent buying on behalf of an application that has no Bloomberg, FactSet, 6sense, or Capital IQ seat and needs a grounded answer in code
The slice does not work if all three of these are true at once: the workflow is dashboard-shaped (scrolling a flow list, scanning a heatmap, watching a feed), free alternatives already serve the non-desk audience well enough (ETF.com, Yahoo Finance, popular Twitter accounts that publish the same info), and the only buyer who sees real uplift over free is a desk analyst already paying for an enterprise contract.Bottom line: “this slice lives inside Bloomberg, FactSet, 6sense, or Capital IQ” is not a disqualifier. The long-term vision is that thousands of unbundled slices, each shipped by an independent contributor, collectively rival those terminals as a decentralized, user-generated intelligence layer for AI agents. What disqualifies a slice is when it lives inside those terminals and has no audience outside them and is inherently dashboard-shaped.

​
Request for Builders by Marketplace Category

These categories match the ones contributors see in the submission form. Use the category that best matches the buyer workflow you are unbundling.

​
Crypto & DeFi

Highest priority right now. This is the strongest current fit for grants because prosumer crypto traders already pay monthly for better market structure, derivatives, and on-chain positioning data, and the workflows map cleanly to natural-language questions an agent would be asked.
Use Grok, Twitter, Reddit, forums, and GitHub issues to find:
  • people complaining they pay for one feature inside an expensive subscription
  • people saying a specific feature is the only reason they keep paying
  • developers complaining about a specific normalization or integration pain
The goal is to isolate one paid workflow, not one broad category.Query example: Sensor Tower users paying mainly for app download estimates or keyword-share data.Execute example: developers losing days normalizing order book or filing schemas across sources.
These are not invitations to clone entire products. They are examples of the specific slices inside premium subscriptions that buyers actually pay for.HyblockCoinGlassGlassnodeBloomberg-style public-data workflows

​
Financial Markets

High priority, with a sharper filter than before. This category is strongest when you unbundle a workflow that a prosumer investor (FinTwit, small-cap community, independent analyst, value investor, short seller) would naturally ask as a question and pay to have answered with evidence. It also works when a slice inside Bloomberg, FactSet, Capital IQ, or 6sense has a meaningful audience outside the desk — prosumer, independent, or agent-driven — even if desks also pay for it today. It does not work when the slice has no audience outside a desk analyst with an enterprise contract. We explicitly support three kinds of Financial Markets RFBs:
  • Prosumer-anchored public-filing workflows: filing deltas, insider clusters, reverse DCF, accounting red flags
  • Bloomberg / FactSet / Capital IQ slices with a non-desk audience: earnings-call language deltas, short-interest and squeeze-risk regime, sell-side rating and target revision velocity, 13F / institutional-ownership deltas, corporate credit and CDS-spread anomalies
  • Narrow developer-facing primitives for normalizing painful public data (filings, ETF disclosures, transcripts)
The classic analyst workflow lines (reverse DCF, earnings quality, comp valuation) still carry a stricter bar because free LLMs already produce a mediocre version. Enterprise-anchored lines also carry a stricter bar: you must show a real non-desk audience or a clear agent-interface advantage over the terminal.
We do not want generic DCF calculators, generic stock screeners, or “analyze any stock” wrappers. Classic analyst workflows are only fundable when they beat a free LLM plus a basic spreadsheet by a meaningful margin. If your pitch’s only buyer is someone with a Bloomberg, FactSet, or Capital IQ seat they won’t cancel and the slice is inherently dashboard-shaped, the line is not a fit. If the same slice also serves prosumer, independent, or agent-driven buyers — or gives desk analysts a natural-language agent interface they would prefer for this specific query — it is.

​
Other

If your idea clearly fits the grant bar but is not Crypto & DeFi or Financial Markets, see the Future Categories (On Ice) section below. We are not currently grant-funding those categories, but the research is preserved there and you can still ship permissionless and earn 90% revenue share.

​
Future Categories (On Ice)

We are not currently grant-funding the categories below because we are focused on bootstrapping Crypto & DeFi and Financial Markets supply first. The research is preserved here because we plan to reopen these as the marketplace expands.You are welcome to ship permissionless in any of these categories today and earn 90% revenue share. If real paying demand shows up on one of these workflows, we will revisit and reopen grants for it.Lines listed as on ice are the ones that survived the agent-native sanity check. Lines that failed it (ETF basis dislocation, Cortellis clones, AI code-remediation) are not listed because they would not convert at $0.10 per answer regardless of grant status. Earnings-estimate / target-revision velocity and Bloomberg-style supply-chain exposure have been reclassified: revision velocity is now funded in the Financial Markets table above, and supply-chain exposure is still off-ramp, but for build complexity reasons rather than framing.
The bar when we reopen this category will be narrow question-shaped workflows for founders and small GTM operators who have no enterprise sales-intelligence budget. Enterprise 6sense / ZoomInfo / BuiltWith customers will not migrate to micropayments, so lines anchored to that audience are not fundable.Not funded even when reopened: generic lead databases, CRM enrichment wrappers, “ZoomInfo for agents,” “all-app” integration gateways, or enterprise-GTM-anchored lines. An authed action is fundable only as one narrow scoped write primitive behind a handshake, not as a broad Zapier-style proxy.
When we reopen this category, ad-spy and app-intelligence workflows survive only if reshaped from “scroll this feed” into “answer this decision question.”Not funded even when reopened: generic SEO suites, broad AI marketing wrappers, “browse this feed of ads” products, or “all-app” integration gateways. Money-moving or write actions are only fundable as one narrow scoped primitive behind a handshake (OAuth and, for spend, a budget envelope), never as a generic Zapier-style proxy.
When we reopen this category, it will only fund tools that give a small operator a decision, not a feed.Not funded even when reopened: weather feeds, sports feeds, news feeds, or location feeds with no decision layer.
When we reopen this category, it will fund two shapes: (1) tools that compress expensive developer toil for solo or small-team devs who have no Snyk / GHAS budget, and (2) narrow agent-infrastructure Execute primitives a coding agent calls directly and consumes raw. The second shape is the lane behind some of the best-selling MCP tools today (web fetch, extraction, component retrieval). It is fundable only when it isolates the specific paid capability, not when it is a generic “call any API” proxy.Not funded even when reopened: generic wrappers around GitHub, Datadog, code-search APIs, or AI code-remediation (Cursor and Claude Code already do this for free with repo context). A scrape or extraction primitive only fits if it isolates the specific paid capability (anti-bot rendering, clean markdown, one structured site class); a generic “fetch any URL” proxy is not fundable.
When we reopen this category, it will only fund tools that beat free paper search for a painful paid workflow.Not funded even when reopened: citation benchmarking (Google Scholar and Semantic Scholar already cover this), clinical-trial site-selection (Cortellis buyers are enterprise biopharma on contracts).
Utility and Other remain very high bar. Most submissions historically drift into generic infrastructure wrappers or dashboard-alerting products that do not belong on Context. When we reopen these, we will list specific fundable slices here. Until then, if you have a strong utility idea that replaces a premium incident, cost, or infrastructure workflow (Datadog, Dynatrace, New Relic, Honeycomb) with a question-shaped buyer, you can still ship permissionless and earn 90% revenue share; contact us if the tool picks up real paid usage and we will consider a retroactive grant.
Important: a good Execute primitive is still narrow. We do not want “all exchange market data” or “all company data.” We want one painful primitive with one stable contract. If you are pitching an SDK-facing primitive, use the category tables above and keep the contract tight.

​
Ideas We Are Less Excited To Fund

We want fewer clever ideas and more fundable products. These patterns usually get rejected or down-ranked:
  • vague AI wrappers
  • “X for agents” terminal clones
  • broad “crypto analytics” or “financial research” platforms
  • social sentiment, meme coin, rug-check, or generic narrative dashboards
  • pure monitoring feeds or alerting products
  • multi-source normalization monsters that a solo builder cannot maintain
  • generic wrappers over public APIs with no schema or evidence advantage
  • products with no obvious paying buyer
  • products better suited to a dashboard SaaS than Context Query or Execute mode
If your pitch sounds like “I will aggregate everything and let the AI figure it out,” it is almost certainly too broad. If your pitch sounds like “I am unbundling this one paid workflow from a product serious users already buy,” you are probably on the right track.

​
How to Apply

One application at a time. If you have an active grant project (Processing, Draft Created, Conversation, or any awarded tier), complete it before submitting a new application. Duplicate submissions will be automatically rejected.Use your GitHub email. Submit your application using the same email address associated with your GitHub account (your public email or the email in your commit history). We verify GitHub ownership automatically. If your application email doesn’t match your GitHub, we’ll ask you to prove you own the account before we can review your proposal.
1

Create Your Account

Sign in at ctxprotocol.com to create your account. This generates your embedded wallet the address where you’ll receive grants and revenue.
Copy your wallet address from the user nav menu in the bottom left of the main chat app. You’ll need it for the application form.
2

Submit Proposal

Fill out our application form with:
  • Your email: use the same email that’s on your GitHub (public profile email or commit email). We verify GitHub ownership automatically, and mismatched emails will delay your review.
  • Your wallet address (from Step 1, required for payment)
  • Tool name and description
  • Exact premium feature or normalized primitive you are shipping
  • Target paying user and why they would pay for this instead of using free substitutes
  • 5 must-win prompts that prove the product contract
  • Expected evidence fields or normalized schema fields
  • Ideal output surface: Query, Execute, or both
  • Freshness / latency target and any ambiguity behavior expectations
  • What expensive service or fragmented integration this replaces
  • Which tier you’re targeting
  • Technical approach (1-2 paragraphs)
  • Your background (GitHub, portfolio)
3

Build & Deploy

Once approved, build your MCP server using our SDK:
Constraint: Tools must return data in under 60 seconds. Focus on instant retrieval and analysis.Follow the Build Tools Guide for the complete pattern.
4

Register Your Tool

Go to ctxprotocol.com/contribute to register your tool:
  • Paste your MCP endpoint URL
  • Add a basic name and description (the optimization skill will generate and push the optimal description in the next step)
  • Set your listing response price to $0 for initial testing (you can change it after we review)
  • Advanced (optional): If you also want SDK developers to call your methods directly with per-call billing (Execute mode), set a default execute price, typically ~1/100 of your response price. Skip this if you’re starting with Query mode only; you can enable it later.
  • Add stake to activate your tool
5

Optimize & Validate

Before requesting review, run the full optimization workflow:Recommended: Load the MCP Tool Optimization Skill into your AI coding agent (Cursor, Claude Code, etc.) with your API key and Tool ID. It researches your vertical, proves your tool beats free LLMs, fixes issues, generates an optimal description, and pushes it to your listing automatically.The skill produces a structured optimization artifact, a machine-readable proof of differentiation. Tier S grants require this artifact (see Quality Requirements below).Also test manually in the Chat app:
  • Enable Developer Mode in Settings → Developer Settings
  • In the right sidebar tool selector turn off Auto mode and select your tool
  • Send a message to query your tool. You should see a response.
  • Check the Developer Logs preview card at the bottom of the response for errors
  • Click Copy All on the logs and paste into an AI assistant to diagnose issues
See the full Debugging Guide for detailed instructions.
Lightweight alternative: For a quick protocol-compliance check without full optimization, use the Deep Validation System Prompt. Use the optimization skill for grant applications, especially Tier S.
6

Request Review

Your tool must be live on the marketplace before requesting review. If you skip the “Contribute to Marketplace” and “Debug & Refine” steps above, your review will be returned and you’ll be asked to complete those steps first.
Once your tool is submitted, tested, and working correctly on the marketplace, email grants@ctxprotocol.com with:
  • Your tool name and endpoint URL
  • Your tool ID from ctxprotocol.com (go to Developer → Tools and copy the ID shown on your tool card). This is how we identify your tool for testing.
  • Link to your public GitHub repo for the MCP server (for code review)
  • 5 must-win prompts that exercise the premium feature or normalized primitive you are claiming
  • Expected evidence fields / expected schema outputs for each prompt
  • Expected response shape for Query (answer_with_evidence and optionally evidence_only) or the exact Execute methods that should pass
  • Your wallet address (for payment verification)
Make sure your listing description is complete. If you used the MCP Server Analysis Prompt in the Contribute step, your description should already include Features, Try Asking questions, and Agent Tips. Reviews that find a poor or generic description will be sent back for improvement.
We’ll test your tool and verify it meets our quality standards before activating grant payments.
7

Get Paid

Once verified, you’re live on the marketplace!
  • Update your listing response price from $0 to your desired response fee
  • Optionally enable execute pricing for methods you want in Execute mode
  • Agents discover and use your tool across Query + Execute
  • You earn 90% of eligible fees in USDC
  • Grant payment follows the schedule below

​
Payment Terms

Grant Payment Schedule (50/50 Split)
  • 50% on passing review: Paid when your tool passes our review (live on marketplace, code reviewed, tests passing)
  • 50% after 30 days: Paid after 30 days of verified uptime and usage
We pay for delivered work. Build and deploy using free infrastructure (Railway free tier, Vercel, etc.), submit for review, and get paid when it passes.

​
Quality Requirements

To receive grant payments, your tool must:
Tier S grants require demonstrated differentiation vs free alternatives. The MCP Tool Optimization Skill produces a structured artifact proving this. Include it with your grant application.

​
Builder Support

You’re not alone. We’re here to help you ship.

GitHub Examples

View reference implementations and code patterns.

​
FAQ

We review proposals within 48 hours. Complex proposals may require a quick call.
Yes, but one at a time. Complete your current grant project before applying for another. This ensures quality and focus.
The RFB is a starting point, not a limit. If your idea is not listed, pitch it using the same contract: exact feature, target user, 5 must-win prompts, expected evidence fields or normalized schema, ideal output surface, and proof that free substitutes are insufficient.
Yes! The grant is a one-time bootstrap. You keep 90% of all eligible usage fees forever that’s the whole point.
All payments (grants and revenue) are in USDC on Base. Low fees, deferred settlement (settled automatically after successful execution).
Yes, as long as you comply with the license. Wrapping an existing API or open-source tool is fine just make it useful for agents.
Tools must return data in under 60 seconds. This isn’t the place for long-running simulations focus on instant data retrieval and analysis.

​
Ready to Build?

The marketplace is hungry for data. AI agents are querying tools right now, and they’ll pay for yours. Think about the data you wish you had access to. The research that costs too much. The insights locked behind enterprise contracts. The analysis you do manually that could be automated. Build it. List it. Get paid.