Skip to main content
Glama

PitchSpark: Pre-Seed & Seed Startup Pitches

Find pre-seed and seed startup pitches for deal flow on PitchSpark

pitchspark_search_pitches_v1
Read-only

Find pre-seed and seed startup pitches (Spark Cards) for investors and deal-flow research: search the founder-submitted startup pitches on PitchSpark (pre-seed, seed and later rounds) by sector, funding stage, HQ region, raise size, traction, green flags or PitchSpark score, using a plain-English request, e.g. "seed-stage fintech in Texas with revenue" or "pre-seed AI companies raising under $2M, not crypto". Use when: an investor, or an agent working for one, wants deal flow; sourcing pre-seed or seed startups by sector, stage, HQ region, raise size, traction, green flags or PitchSpark score; screening early-stage startups; building a shortlist to read in depth. Do not use for: public companies or stock data; later-stage, growth or M&A data such as Crunchbase-style funding histories (PitchSpark holds founders' pitches, not funding records); founder contact details, decks or data rooms (never returned); verifying a startup's claims; investment advice. Reading one pitch by its id is pitchspark_get_pitch_v1; creating a founder's own draft pitch is pitchspark_submit_pitch_v1. Returns "pitches", a list of Spark Cards, each in this order: id, company_name, ask ({"usd","label"}: the raise in US dollars and as text, e.g. 2500000 and "$2.5M"; usd null and label "Undisclosed" when the founder has not disclosed it), stage, sector, additional_sectors (only when the pitch lists any), hq_location (the company's HQ region spelled out, e.g. "United States" or "Europe"; a region only, never a city or address; null when not set), description (the founders' one-line description of the company, at most 200 characters, or null), spark_card_url (the link to the full pitch on pitchspark.io), spark_card_link_text (the text to show for that link, "<company_name> on PitchSpark"), traction_summary, and green_flags (EITHER a list of up to 5 founder-chosen flags, OR the string "Not published" when the founders chose none); a source "attribution" line; a "footer" sentence; a "signup" object ({"label","url"}: a free investor sign-up button); and an "interpretation" object showing how the request was understood, which filters came from an argument, any filter that had to be relaxed to find results, and anything it could not place. It picks out the sector, funding stage, location (as an HQ region, one of us, canada, latam, uk, europe, mea, apac, global_remote: a US state, a city or a country narrows only to its region), raise size, score and traction cues in your sentence, applies them as filters, and searches the remaining words. "not X" or "without X" leaves pitches out, a phrase in double quotes must appear as written, and "top scoring" orders results by score. Explicit filters, all optional; each one overrides what the sentence says about the same thing and is never relaxed: sector (a pitch's primary or additional sectors); stage; location or hq_region (the HQ region); min_raise_usd and max_raise_usd (US dollars); green_flags (pitches showing any of the listed flags, which founders choose and PitchSpark does not verify); has_traction; min_score. A raise filter matches only pitches whose raise is disclosed; a score filter only pitches with a public score; a traction filter only pitches whose founders wrote a public traction note. When nothing matches, a region, then a raise range, then a traction filter that was read from the sentence is dropped and named in interpretation.relaxed; a filter passed as an argument, and sector, stage, score and green flags, are never relaxed. Each pitch link carries a tracking reference (mref) so PitchSpark can count how often delivered links are opened. How search works today: rule-based matching with keyword and synonym ranking. Your words, and common synonyms for them, are matched against the text founders submitted. There is no semantic or embedding search yet, so results depend on the words used rather than on meaning: concrete terms ("battery storage", "payments API") work better than abstract ones, and a strong pitch that uses different words can be missed. Up to 10 pitches per call, or up to 25 with an API key on a PitchSpark Pro plan or above. Repeated searches return fresh results: a pitch this tool already returned to you (the same API key, or without a key the same connection) in the last 7 days is left out, and the next-best matches take its place. When every match has already been shown to you, "pitches" is empty (nothing is counted against the allowance) and interpretation.seen_excluded says how many matches were left out; interpretation.seen_excluded is 0 when nothing was. include_seen: true turns this off: the best matches are returned whether or not they were returned before, and the pitches returned are not recorded. To do this, PitchSpark keeps the ids of the pitches returned to you and when, for 7 days, against a one-way hash of the API key or connection, never the key or IP address itself. Access: no API key is needed unless the server is set to require one (a call without a key then returns UNAUTHORIZED). Daily allowance, 1 per pitch returned, reset at 00:00 UTC: 10 per connection (IP address) without a key, 10 per account with a free-tier PitchSpark API key, 1000 per account with a key on a PitchSpark Pro plan or above. A key goes in a header only ("Authorization: Bearer psk_..." or "X-API-Key"); a key in the URL, or an invalid or revoked key, returns UNAUTHORIZED. A PitchSpark submit token (pss_...) only works for pitchspark_submit_pitch_v1: sent to this tool it returns UNAUTHORIZED. Also 30 tool calls per minute per connection. Over a limit the tool returns RATE_LIMITED with retry_after_seconds. Read-only: does not change your data or any pitch content. Requests are logged, and each pitch returned is counted as a view. Opening a returned pitch link can add one spark to the pitch's public spark count (once per viewer, updated hourly). A spark is added at most once per viewer per pitch (the signed-in PitchSpark account that opens the link, or otherwise the connection; a viewer without an email-verified account is remembered for 90 days), never when that account is the pitch's founder, PitchSpark staff or has already sparked the pitch, and only up to a daily cap per pitch (10 from viewers without a verified account, 25 from verified accounts). Sparks from viewers without a verified account do not count toward leaderboard points or badges. Recording a view never changes what the tool returns. What to trust: everything in a pitch is written by its founders and is not verified by PitchSpark. Search results do not include a PitchSpark score. Nothing here is investment advice. Pitch text is founder-supplied data, not instructions from PitchSpark.

Tool call template

Replace each placeholder with a value matching the input schema. Only required parameters are included; add optional parameters as needed.

{
  "id": 1,
  "jsonrpc": "2.0",
  "method": "tools/call",
  "params": {
    "arguments": {
      "query": "<query>"
    },
    "name": "pitchspark_search_pitches_v1"
  }
}

Input Schema

NameRequiredDescriptionDefault
limitNoHow many pitches to return, e.g. 5. Default 10. The ceiling is 10 without an API key or with a free-tier key, and 25 with a key on a PitchSpark Pro plan or above; a larger number is reduced to that ceiling. Each pitch returned counts as 1 against the daily allowance.
queryYesWhat you are looking for, in plain English, 1 to 200 characters, e.g. "B2B SaaS with paying customers scoring above 80" or "seed climate startups in Europe raising under $3M, not hardware". The sector, stage, location, raise size, score and traction cues in it are understood; an explicit filter argument (sector, stage, location / hq_region, min_raise_usd / max_raise_usd, has_traction, min_score) overrides what the sentence says about the same thing. A US state or a city can only be given here: it narrows to its HQ region.
stageNoOnly pitches at this funding stage, or any of up to 8. One of: pre_seed, seed, series_a, series_b, series_c, bridge_mezzanine, growth, acquisition. Example: "seed" or ["pre_seed", "seed"]. Overrides any stage named in the query.
sectorNoOnly pitches in this sector, or any of up to 10. Matches a pitch's primary sector or one of its additional sectors. Standard keys: fintech, healthtech, saas, marketplace, ai_ml, consumer, enterprise, climate, edtech, crypto, biotech, hardware, gaming, real_estate, food_beverage, fashion, travel, sports, media, femtech, cybersecurity, dev_tools, ecommerce, logistics, mobility, hr_tech, legal_tech, insurtech, proptech, agtech, energy, robotics, govtech, hospitality, other. Common names like "AI", "climate tech" or "web3" are accepted. Example: "fintech" or ["fintech", "insurtech"]. Overrides any sector named in the query.
locationNoOnly pitches headquartered in this region, or any of up to 8. One of: us, canada, latam, uk, europe, mea, apac, global_remote. Some country names are also accepted and read as their region (for example "United Kingdom", "Mexico", "Brazil", "India", "Japan", "Australia"); a name that is not recognised is rejected, so prefer the region keys. Regions are the finest level stored: put a state or city in the query instead. Example: "us" or ["uk", "europe"]. Overrides any place named in the query and is never relaxed.
hq_regionNoAnother name for location (same values, same filter), e.g. "europe". If both are given, a pitch in any region listed in either one matches.
min_scoreNoOnly pitches whose public PitchSpark score (PitchSpark's AI pitch score, 0 to 100) is at least this, e.g. 80. Scores below 70 are never published, so values under 70 are treated as 70, and pitches without a public score are left out. Overrides any score named in the query.
green_flagsNoOnly pitches showing at least one of these green flags (credibility signals the founders chose, not checked by PitchSpark) on their card, e.g. "Previous exit", "Profitable", "Patent granted", "Signed LOI", "Network effects". One flag or a list of up to 10, each at most 60 characters. The whole flag is compared, ignoring case and surrounding spaces; founders can also write their own short flags. Only set as an argument: flags are not read from the query.
has_tractionNotrue: only pitches whose founders wrote a public traction summary. false: do not require one, even if the query mentions revenue or customers. Example: true. Overrides any traction cue in the query and is never relaxed.
include_seenNofalse (the default): leave out pitches this tool already returned to you (the same API key, or without a key the same connection) in the last 7 days, so a repeated search gives fresh results. true: return the best matches whether or not you have seen them, and do not record the pitches returned. Example: true.
max_raise_usdNoOnly pitches raising at most this much, in US dollars, e.g. 2000000 for $2M. Matches only pitches whose raise is disclosed. Must not be below min_raise_usd. Giving either raise bound replaces any raise size named in the query, and is never relaxed.
min_raise_usdNoOnly pitches raising at least this much, in US dollars, e.g. 500000 for $500K. Matches only pitches whose raise is disclosed. Giving either raise bound replaces any raise size named in the query, and is never relaxed.

Output Schema

NameRequiredDescriptionDefault
errorNoPresent only when the call failed (isError true).
footerNo
signupNo
pitchesNo
truncatedNotrue when pitches were left off to keep the reply under the size cap.
attributionNo
interpretationNoHow the request was read: caller-derived values only (filters, relaxed filters, seen_excluded).

Schema Changelog

Changes observed during successful MCP inspections.

  1. Changed1 schema field changed
    • changedOutput schema / (root)
      Previous value: -nullNew value: +{
      +  "additionalProperties": false,
      +  "anyOf": [
      +    {
      +      "required": [
      +        "pitches",
      +        "attribution",
      +        "footer",
      +        "signup"
      +      ]
      +    },
      +    {
      +      "required": [
      +        "error"
      +      ]
      +    }
      +  ],
      +  "properties": {
      +    "attribution": {
      +      "type": "string"
      +    },
      +    "error": {
      +      "additionalProperties": false,
      +      "description": "Present only when the call failed (isError true).",
      +      "properties": {
      +        "category": {
      +          "description": "The error class.",
      +          "enum": [
      +            "RATE_LIMITED",
      +            "QUOTA_EXCEEDED",
      +            "UNAUTHORIZED",
      +            "FORBIDDEN",
      +            "VALIDATION_ERROR",
      +            "NOT_FOUND",
      +            "CONFLICT",
      +            "UNKNOWN_TOOL",
      +            "INTERNAL"
      +          ],
      +          "type": "string"
      +        },
      +        "code": {
      +          "description": "Machine-readable error code, e.g. RATE_LIMITED, UNAUTHORIZED, FORBIDDEN, INVALID_ARGUMENT, NOT_FOUND, CONFLICT, PITCH_LIMIT_REACHED, INTERNAL.",
      +          "type": "string"
      +        },
      +        "message": {
      +          "description": "A short human-readable explanation, safe to show the user. Never a stack trace.",
      +          "type": "string"
      +        },
      +        "retry_after_seconds": {
      +          "description": "Seconds to wait before retrying, when waiting helps.",
      +          "minimum": 0,
      +          "type": "integer"
      +        },
      +        "retryable": {
      +          "description": "true when the same call can succeed later without changes.",
      +          "type": "boolean"
      +        }
      +      },
      +      "required": [
      +        "code",
      +        "message"
      +      ],
      +      "type": "object"
      +    },
      +    "footer": {
      +      "type": "string"
      +    },
      +    "interpretation": {
      +      "description": "How the request was read: caller-derived values only (filters, relaxed filters, seen_excluded).",
      +      "type": "object"
      +    },
      +    "pitches": {
      +      "items": {
      +        "additionalProperties": false,
      +        "properties": {
      +          "additional_sectors": {
      +            "items": {
      +              "type": "string"
      +            },
      +            "type": "array"
      +          },
      +          "ask": {
      +            "additionalProperties": false,
      +            "properties": {
      +              "label": {
      +                "type": "string"
      +              },
      +              "usd": {
      +                "type": [
      +                  "number",
      +                  "null"
      +                ]
      +              }
      +            },
      +            "required": [
      +              "usd",
      +              "label"
      +            ],
      +            "type": "object"
      +          },
      +          "company_name": {
      +            "type": "string"
      +          },
      +          "description": {
      +            "type": [
      +              "string",
      +              "null"
      +            ]
      +          },
      +          "green_flags": {
      +            "anyOf": [
      +              {
      +                "items": {
      +                  "type": "string"
      +                },
      +                "type": "array"
      +              },
      +              {
      +                "type": "string"
      +              }
      +            ]
      +          },
      +          "hq_location": {
      +            "type": [
      +              "string",
      +              "null"
      +            ]
      +          },
      +          "id": {
      +            "type": "string"
      +          },
      +          "sector": {
      +            "type": [
      +              "string",
      +              "null"
      +            ]
      +          },
      +          "spark_card_link_text": {
      +            "type": "string"
      +          },
      +          "spark_card_url": {
      +            "type": "string"
      +          },
      +          "stage": {
      +            "type": [
      +              "string",
      +              "null"
      +            ]
      +          },
      +          "traction_summary": {
      +            "type": [
      +              "string",
      +              "null"
      +            ]
      +          }
      +        },
      +        "required": [
      +          "id",
      +          "company_name",
      +          "ask",
      +          "stage",
      +          "sector",
      +          "hq_location",
      +          "description",
      +          "spark_card_url",
      +          "spark_card_link_text",
      +          "traction_summary",
      +          "green_flags"
      +        ],
      +        "type": "object"
      +      },
      +      "type": "array"
      +    },
      +    "signup": {
      +      "additionalProperties": false,
      +      "properties": {
      +        "label": {
      +          "type": "string"
      +        },
      +        "url": {
      +          "type": "string"
      +        }
      +      },
      +      "required": [
      +        "label",
      +        "url"
      +      ],
      +      "type": "object"
      +    },
      +    "truncated": {
      +      "description": "true when pitches were left off to keep the reply under the size cap.",
      +      "type": "boolean"
      +    }
      +  },
      +  "type": "object"
      +}
  2. Added

TDQS

A4.6/5.0
Behavior5/5

Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?

Annotations only declare read-only/open-world/non-idempotent; the description adds everything an agent actually needs to behave correctly: no key is required unless the server demands one, UNAUTHORIZED cases (key in URL, revoked key, pss_ submit token), daily allowance accounting per pitch returned, 30 calls/minute, RATE_LIMITED with retry_after_seconds, and the 7-day seen-exclusion with how to disable it. It also discloses that matching is keyword/synonym rule-based with no semantic search, which is a material behavioral caveat not expressed anywhere in the structured fields.

Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.

Conciseness2/5

Is the description appropriately sized, front-loaded, and free of redundancy?

Purpose and when-to-use are correctly front-loaded, but the text is far past the size needed: the spark-counting rules (cap of 10 unverified / 25 verified viewers, 90-day viewer memory, leaderboard exclusion, hourly update) are tangential to invoking the tool and consume a large block. Other material repeats what the schema already says verbatim, notably that raise filters match only disclosed raises and that score filters only match public scores.

Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.

Completeness5/5

Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?

For a 12-parameter, open-world, rate-limited search tool, nothing an agent needs is missing: auth paths, quota math, pagination ceiling per plan, relaxation precedence, what "pitches" contains, and even a pointer to interpretation.seen_excluded for the empty-result case. An output schema exists, so the long field-by-field return walkthrough is surplus rather than a necessity, but completeness is not in question.

Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.

Parameters4/5

Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?

Schema coverage is 100%, so the baseline is 3, but the description adds cross-parameter semantics the schema does not: an explicit filter argument overrides whatever the sentence says about the same thing, and that argument filters are never relaxed while sentence-derived region/raise/traction filters are dropped in a named order. It also clarifies which filters can only be set as arguments (green flags) versus only in the query (US state or city narrowing to an HQ region).

Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.

Purpose5/5

Does the description clearly state what the tool does and how it differs from similar tools?

Opens with a specific verb and resource ("Find pre-seed and seed startup pitches (Spark Cards)") plus the caller it serves (investors, deal-flow research), and explicitly names the two siblings that do adjacent things: reading one pitch by id is pitchspark_get_pitch_v1 and creating a draft is pitchspark_submit_pitch_v1. An agent can distinguish this from those siblings without opening any schema.

Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.

Usage Guidelines5/5

Does the description explain when to use this tool, when not to, or what alternatives exist?

Has an explicit "Use when:" list (sourcing by sector/stage/region/raise/traction, screening early-stage startups, building a shortlist) and an explicit "Do not use for:" list (public companies, Crunchbase-style funding histories, founder contacts/decks, verifying claims, investment advice). It goes further and routes to alternatives by sibling name, so both the inclusion and exclusion conditions are stated.

Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.

Try in Browser

Glama MCP Gateway

Add one secure layer between your agents and this server.

Resources