Momus Agent Marketplace

npx add-mcp 'https://market.momusdigital.com/mcp'

Agent guide and offer contracts

Public discovery is free. Purchases require your operator’s authorized budget and hosted payment.

Read current availability and terms before preparing a result.

Complete plain-text agent guide · OpenAPI · JSON catalog · Open the storefront

Guest access, preparation and payment

1. Inspect the exact offer input/output contract, price, source limitations and terms. Obtain operator authorization for the wanted purchase.
2. Use HTTP POST /api/guest with an empty JSON object. No email or password is required. Save the returned access token privately; this bootstrap happens outside MCP. Configure Authorization: Bearer <access_token> for private API or MCP calls at /mcp. Never put the token in a URL or shared log.
3. Submit the exact input, accepted terms and one persistent idempotency key to POST /api/checkout or MCP create_checkout. Repeat unchanged arguments while preparation is running. Busy or failed preparation opens no payment session.
4. Only after preparation succeeds, hand the hosted Stripe Checkout URL to the authorized operator. The service does not autonomously charge a card. A browser return URL is not payment evidence.
5. Reconcile the existing order. Canonical verified payment unlocks that same prepared artifact; retrieve it with the same private access and check its receipt and output contract.

Downloads expire 30 days from completion. Payment creates no future run, subscription or hardware reservation. Guest access can be saved privately or optionally converted to an account; see the complete guide for recovery and client setup.

Current terms: 2026-10-03-immediate-1. All sales are final. We do not offer discretionary refunds. This policy does not limit rights or remedies that applicable law or payment-network rules require. Failed delivery and payment disputes remain reviewable service obligations.

17 bounded offers

Examples and source snapshots are documentation, not customer purchases. Game metadata does not establish verified crossplay. Texas snapshots are not current ticket inventory or predictions of winnings. Generated outputs require review.

Game catalog metadata lookup

$1.00 USD per prepared result

Find recorded game titles, platform editions, multiplayer flags and region links in the owned CrossPlatGames copy of LaunchBox metadata. The imported database contains 179,897 rows across 189 platform names; rows include platform editions and are not 179,897 distinct games. Use it to identify catalog records for further verification, not to decide whether two editions support crossplay.

Input: query is a literal title substring (up to 100 characters); limit is 1–25, default 10. SQLite LIKE matching is ASCII case-insensitive, with literal percent/underscore characters; no fuzzy matching. A blank query matches initial records sorted by title then ID.

Output: JSON with query, returned count, limit, game records, recorded platforms/regions, source provenance, matching rule and scope. No artwork, ROMs or game binaries are delivered.

Availability: Available when bounded capacity permits.

Read the Game catalog metadata lookup JSON contract

Source: Owned CrossPlatGames / LaunchBox SQLite metadata; imported with owner approval. · Captured 2026-10-03T09:40:13Z; source file modified 2026-05-20T04:01:53Z; underlying dataset as-of date is unknown.

Catalog metadata does not establish cross-play compatibility, online server availability, edition equivalence or current store availability. Source dates are retained; no ROMs or images.

• 179,897 platform-edition rows, 189 platform names; not a deduplicated game count.
• Multiplayer/cooperative flags are not verified crossplay evidence.
• No current storefront, server, matchmaking or regional availability guarantee.

Texas scratch-off snapshot

$1.00 USD per prepared result

Retrieve source-labeled Texas scratch-off game data from a fixed owned snapshot: 74 game numbers, 74 official views and 734 official prize-tier records, with separate modeled context. Use it for reproducible data inspection or analysis with explicit dates. This offer does not sell lottery tickets or identify tickets to buy.

Input: Optional query matches a literal case-insensitive substring of title/name/game number/ID; blank returns initial source records. Optional limit is 1–25, default 10. Matching preserves the source order and never fetches current inventory.

Output: JSON with matching_count before truncation, returned count, source game summaries, official tier objects and modeled_context, provenance dates, scope and field_note. Source fields retain their original names; latest_estimated/latest_modeled/profit_odds values are model outputs, distinct from official counts.

Availability: Available when bounded capacity permits.

Read the Texas scratch-off snapshot JSON contract

Source: Owned TXLotteryOdds API snapshot; 150 archived responses covering catalog, official views and modeled views. · 2026-10-03 snapshot date for all 74 games; fixed until deliberately refreshed.

Current available data is incomplete and may be stale. Historical information only; no ticket sales, retailer stock claims or prediction of winning tickets. Modeled estimates remain labeled.

• Six source games carry anomaly flags; they remain visible.
• Public-site parity and current freshness are unverified.
• Modeled estimates can be incomplete or wrong; they do not predict a particular ticket.
• Artwork URLs are metadata only; no artwork bytes or rights are supplied.

Testable task recipes

$1.00 USD per prepared result

Get a fixed, authored procedure for a common agent task, including ordered steps, acceptance checks, reference URLs and a strongest free/direct starting point. Choose this when the missing piece is a reproducible method, not execution of the task.

Input: Choose exactly one topic: csv-to-json, gltf-validation, webhook-idempotency, local-model-json or ffmpeg-social.

Output: JSON containing topic, title, steps, acceptance, free_baseline, sources and verification. Some topics include an example or example_command. References are identified but not fetched and commands are not executed.

Availability: Available when bounded capacity permits.

Read the Testable task recipes JSON contract

Source: Original procedural guidance authored in this marketplace implementation, with named primary reference URLs. · 2026-10-03 recipe version.

Recipes are bounded guidance. Tool versions and the actual output must be checked in your environment.

• No tools run or source pages are fetched for this purchase.
• A recipe is not evidence that your task has been completed.

CSV → structured JSON

$1.00 USD per prepared result

Convert a small comma-separated UTF-8 text table into JSON records without losing duplicates, quoted commas or embedded newlines. Use it when a downstream agent needs every cell preserved as text and malformed rows reported as a visible failure.

Input: csv includes exactly one header row and up to 1,000 data rows, 50 columns and 100,000 input characters. Headers must be unique, nonblank and at most 128 characters. Comma delimiter and doubled-quote escaping only.

Output: JSON headers, rows, row_count, column_count and value_policy. Every cell is a string, including numbers, empty cells and dates; no inferred types or silent repair. Malformed quotes/row widths and excessive expansion reject preparation before checkout.

Availability: Available when bounded capacity permits.

Read the CSV → structured JSON JSON contract

Source: Only the CSV text supplied by the requester; no external fetch. · Request input.

UTF-8 text; comma delimiter; at most 1,000 data rows, 50 columns and 128 characters per header. Inputs that expand beyond the bounded JSON result budget are rejected. No silent row dropping or type guessing.

• No type inference, delimiter detection, spreadsheet evaluation or source fact verification.

Exact-key row deduplication

$1.00 USD per prepared result

Keep the first record for each exact scalar key while retaining a ledger of every discarded duplicate and whether it conflicts. Useful for reproducible deduplication where silent merging would hide data loss.

Input: rows contains at most 1,000 objects with at most 50 properties each. key names an own property present in every row; values must be strings, finite numbers, booleans or null. Type is part of identity: numeric 1 and string "1" differ.

Output: JSON with retained rows in first-seen order, duplicates including original/kept zero-based row indices and discarded rows, conflict flags, and input/kept/duplicate counts.

Availability: Available when bounded capacity permits.

Read the Exact-key row deduplication JSON contract

Source: Requester-supplied rows and exact key field. · Request input.

Exact scalar keys only. Conflicting duplicates are reported; semantic entity matching is outside scope.

• No fuzzy entity matching, normalization or semantic merge.

Source-linked brief

$1.00 USD per prepared result

Turn supplied factual statements and HTTP(S) source links into an escaped Markdown evidence brief and a portable source ledger. Useful for packaging already collected evidence without implying that research was independently performed.

Input: title is at most 120 characters. Supply 1–20 sources, each with a URL up to 500 characters and a fact up to 2,000 characters. URLs must be HTTP(S), absolute and have no embedded credentials.

Output: JSON title, normalized sources and Markdown; ZIP download contains brief.md and sources.json. Each source is labeled requester-supplied and not independently verified.

Availability: Available when bounded capacity permits.

Read the Source-linked brief JSON contract

Source: Only requester-supplied facts and URLs. · Request input; underlying source freshness unknown.

Formats supplied statements without independently verifying or fetching their sources. No invented research or endorsement.

• No browsing, citation validation, fact checking or endorsement.

Event-history gap audit

$1.00 USD per prepared result

Inspect exported block-number observations for numeric gaps, repeated block numbers and conflicting supplied hashes. Use it to identify rows that need explanation before further event-history analysis.

Input: events is 1–1,000 objects with a nonnegative safe-integer block number and optional hash up to 100 characters. Input order need not be sorted; missing spans are summarized as ranges, not expanded.

Output: JSON counts, first/last block, missing_block_count and ranges, duplicate block occurrences with zero-based row indices, conflicting_hashes and an interpretation notice.

Availability: Available when bounded capacity permits.

Read the Event-history gap audit JSON contract

Source: Requester-supplied event export only; no chain connection. · Request input.

Read-only analysis of your export. It cannot establish chain finality or whether blocks without relevant events should be present. No trading or transactions.

• No finality, consensus, missing-event proof, trading or transactions.

Original SVG interface kit

$2.00 USD per prepared result

Generate six original editable line icons—search, check, download, grid, bolt and document—with your color and stroke width. Use the ZIP as a small interface starting kit, not a custom brand identity.

Input: color is exactly a six-digit hexadecimal color prefixed by #. stroke is optional, integer 1–3, default 2. Shapes and 24×24 viewBox are fixed.

Output: JSON icon names/count, color, stroke and original-source provenance. ZIP contains six.svg files, tokens.json, manifest.json, README.md and the frozen purchase LICENSE.txt.

Availability: Available when bounded capacity permits.

Read the Original SVG interface kit JSON contract

Source: Fixed original vector geometry generated by this implementation; no third-party stock. · Generated from this request.

Procedural, original vector geometry. No third-party stock, logos, trademarks or claim of exclusivity.

• Nonexclusive; no logos, trademark rights or bespoke design service.

Procedural glTF starter

$2.00 USD per prepared result

Generate an editable one-meter cube centered at the origin in glTF 2.0 with embedded geometry, flat normals and a solid-color material. Use it as a predictable import fixture or starting mesh.

Input: color is # plus six hexadecimal digits; optional name is at most 60 characters, default Momus starter cube. The input sRGB color is converted to a linear material factor.

Output: JSON manifest with format, dimensions, 24 vertices, 12 triangles, embedded-buffer flag and validation scope. ZIP includes cube.gltf, manifest.json, README.md and the frozen LICENSE.txt; no external buffer/image fetch is required.

Availability: Available when bounded capacity permits.

Read the Procedural glTF starter JSON contract

Source: Original deterministic cube construction with embedded bytes; no stock asset. · Generated from this request.

A simple procedural mesh, not a commissioned production scene. Import and material behavior still depend on the target engine.

• No custom scene, rigging, textures, animation or engine-specific compatibility guarantee.

Accessible page starter

$2.00 USD per prepared result

Generate a self-contained editable HTML/CSS page with navigation, hero text, cards, contact placeholder, semantic landmarks and keyboard-focus styling. Use it as a static project starter that you will edit and validate before publishing.

Input: Supply a title up to 100 characters and description up to 500 characters. They are inserted as escaped text. Feature cards and contact content are placeholders.

Output: JSON file list and accessibility scope; ZIP contains index.html, styles.css, README.md, manifest.json and the frozen LICENSE.txt. It uses local CSS and system fonts, with no JavaScript, analytics or external font dependencies.

Availability: Available when bounded capacity permits.

Read the Accessible page starter JSON contract

Source: Original local HTML/CSS template; requester-supplied title and description. · Generated from this request.

Static source files; no hosting, backend, tracking pixels or external fonts included.

• No hosting, backend, domain, contact delivery or complete accessibility audit.

Synthetic record fixtures

$1.00 USD per prepared result

Create reproducible synthetic records for application tests, with a deterministic seed and an accompanying JSON contract. Useful for test setup when real personal or transaction data is unnecessary.

Input: count is 1–200; seed is an integer from 1 to 2,147,483,646. The same values produce the same records.

Output: JSON rows with id/name/example.invalid email/score/active/synthetic fields, provenance and field contract; ZIP contains fixtures.json, contract.json, manifest.json and the frozen LICENSE.txt.

Availability: Available when bounded capacity permits.

Read the Synthetic record fixtures JSON contract

Source: Park–Miller generator (16807, modulus 2147483647); no personal records. · Generated from this request.

Clearly labeled synthetic records for software tests. They are not real people, transactions, market evidence or training benchmarks.

• Not real people, transactions, market demand or a validated training benchmark.

JSON field-contract check

$1.00 USD per prepared result

Check a flat required-field/type contract over supplied JSON objects and return visible violations. Use it before passing records to a downstream tool that requires predictable field types.

Input: rows contains up to 1,000 objects with up to 50 properties. fields maps 1–30 required names (1–128 characters) to string, number, boolean, array, object or null. Extra row fields are allowed.

Output: JSON valid flag, row_count, submitted fields, total violation_count, at most 100 errors with zero-based row_index/field/expected/actual, and errors_truncated.

Availability: Available when bounded capacity permits.

Read the JSON field-contract check JSON contract

Source: Requester-supplied records and field/type mapping. · Request input.

A flat field/type contract with field names of 1–128 characters, not full JSON Schema or a claim of semantic correctness. Reports at most 100 example violations plus the total.

• No nesting rules, coercion, value ranges or full JSON Schema validation.

Row-loss reconciliation

$1.00 USD per prepared result

Compare expected and delivered ID multisets to reveal missing IDs, unexpected IDs and duplicate multiplicity. Useful as an acceptance check after a data pipeline or export.

Input: expected and actual are arrays of at most 1,000 exact strings, each up to 100 characters. Case, whitespace and duplicate occurrence counts matter; ordering does not.

Output: JSON valid flag, total expected/actual counts, missing/unexpected entries with deficit/excess counts, multiplicity ledger and matching rule.

Availability: Available when bounded capacity permits.

Read the Row-loss reconciliation JSON contract

Source: Requester-supplied expected and actual ID lists. · Request input.

Checks identities and multiplicity. Does not verify the contents of each row.

• No row-content, ordering, fuzzy-match or source-completeness verification.

Workflow dependency check

$1.00 USD per prepared result

Validate a supplied task dependency graph for repeated IDs, missing/duplicate dependencies and cycles; return a dependency-first execution order when valid. Use it to check a plan before your own executor runs anything.

Input: tasks contains 1–100 objects. Each has an id up to 60 characters and depends_on with at most 30 IDs. Every dependency must identify a supplied task.

Output: JSON diagnostics and total counts, per-category truncation flags, cycle witnesses, valid flag and execution_order (null if invalid). executed is always false. Diagnostic examples cap at 100 per category and cycle examples at 128 KiB.

Availability: Available when bounded capacity permits.

Read the Workflow dependency check JSON contract

Source: Requester-supplied task graph only. · Request input.

Checks the submitted graph only. Diagnostic examples are capped at 100 per category, with a 128 KiB cycle-example budget and total counts. Does not execute tasks or guarantee external services are available.

• No scheduling, execution, retries, external services or semantic dependency inference.

Media delivery manifest check

$1.00 USD per prepared result

Check supplied media delivery metadata for safe relative paths and compute a sequential timeline from declared clip durations. Useful for validating a handoff manifest before inspecting or rendering actual media.

Input: clips contains 1–100 entries with a relative POSIX path up to 200 characters and duration_seconds greater than 0 and at most 86,400. Optional caption_path is also relative. Absolute paths, traversal, schemes and backslashes are rejected in the report.

Output: JSON valid flag, clip_count, cumulative total duration, timeline with start/end seconds, errors and warnings. Repeated clip paths and unusual caption extensions warn; bytes_inspected is always false.

Availability: Available when bounded capacity permits.

Read the Media delivery manifest check JSON contract

Source: Requester-supplied manifest; no media bytes accessed. · Request input.

Metadata validation only. Does not inspect media bytes, speech, frame accuracy or whether referenced files exist.

• No rendering, file-existence check, audio/video inspection or actual caption validation.

URL input contract

$1.00 USD per prepared result

Perform offline URL syntax screening and flag embedded credentials, local/private address literals, unsupported schemes, nonstandard ports and whitespace/control characters. Use it as one input-contract check, never as a complete network security boundary.

Input: urls contains 1–100 strings of at most 1,000 characters. HTTP(S) URLs are considered; normalization uses the runtime URL parser. No DNS request or destination fetch occurs.

Output: JSON overall valid flag and one result per input with flags; parseable URLs add normalized URL, scheme and hostname. dns_resolved=false and network_requests=0 explicitly bound the check.

Availability: Available when bounded capacity permits.

Read the URL input contract JSON contract

Source: Requester-supplied strings and local URL parsing only. · Request input; destination state unknown.

Lexical input screening only; no DNS resolution. It is not a complete SSRF defense or a destination trust assessment.

• No DNS resolution, redirects, TLS/reputation checks or protection against rebinding.

Bounded local-model answer

$3.00 USD per prepared result

Generate one bounded text answer using the owned Qwen 3.8 27B profile, for a short drafting, explanation or transformation task. The model receives only your text and has no browser, tools or authority to act on external systems.

Input: prompt is at most 12,000 characters. Optional max_tokens is 64–1,024, default 256. The request has a 120-second deadline and only starts when capacity/model readiness permits.

Output: JSON answer, reported model identity, available token usage, elapsed_ms, finish_reason and requested_max_tokens. tools_used=false and output_needs_review=true; the answer may be truncated or wrong.

Availability: Available when bounded capacity permits.

Read the Bounded local-model answer JSON contract

Source: Owned qwen3.8-27b-nvfp4 profile through a text-only broker; requester prompt. · Generated at preparation time; no live retrieval.

Local inference, with a 120-second deadline and 1,024 output-token maximum. Model output needs review. No browser, tools, hosted fallback or background agent access.

• Model output can hallucinate, omit requirements or be truncated.
• No hosted fallback, future work, reservations, tool execution or guaranteed correctness.