Instructions for coding agents

Copy the block for the way you’re integrating Pearch. Each block includes the integration steps, search recommendations, and references the agent can use.

API integration

FOLLOW THIS SETUP ORDER
1. The user creates a Pearch API key and stores it in backend secret configuration. Never ask them to paste it into chat.
2. Inspect the stack. Install the official SDK only for a Python integration, or configure an HTTP client for another stack.
3. Call GET /v1/user (PearchClient.get_user() in Python) to verify the account for free. Keep credentials out of output. Report the account, credits and prices.
4. Prepare a small people search with reviewed search_requirements. Ask for approval before running it because it uses Pearch credits.

Help me integrate Pearch into this product. First inspect the existing stack and user flow. Ask for any missing product decisions. Follow the attached guide and current API documentation. Prefer search_requirements with human review. Keep the API key on the server or in a secure runtime, never in browser code or the prompt. Use the official Python SDK if this is a Python integration. Apply the recommendations to both backend behavior and the user interface. Flag any deviations in the existing product to the developer, with the impact and a suggested fix. Confirm unresolved product decisions before implementation. Implement the agreed flow. Do not make paid production searches without my approval.

PRODUCT RECOMMENDATIONS

## Choose how users provide search intent

Input quality has the strongest influence on search quality. Good input is specific, structured, observable on a professional profile, and free of contradictions.

The hiring manager is the source of truth. Before searching, they should approve the target profile and business context, 2–6 real must-haves, nice-to-haves, and exclusions. A recruiter, integration, or LLM can structure the request, but should not silently choose essential criteria or trade-offs.

- Requirements first (recommended): Users enter structured search_requirements and review them before searching. They can see each criterion, mark must-haves and nice-to-haves, and resolve conflicts before spending credits. This gives them the most control.

- Query, review, then rerun: Users start with a natural-language query. Pearch converts it into search_requirements. Show those requirements for human review, then rerun the search with the approved criteria. This balances a simple input experience with explicit control.

- Query only: The search runs without a review of the interpreted requirements. This is supported but offers less control: vague, unstructured, or contradictory queries can lose important distinctions.

## Write clear, verifiable requirements

Each requirement should express one clear idea and be verifiable from a professional profile.

Avoid requirements that cannot normally be verified from a professional profile, such as willingness to accept a salary, availability to start on a specific date, or being a “team player.”

- Must-haves and preferences: Use 2–6 must-haves. A criterion is mandatory only if the customer would reject a candidate without it. Keep preferences as nice-to-haves so they improve ranking without excluding relevant candidates.

- Connected context: Keep a skill and how it was used together, or a company and the required period. These connected conditions belong in the same requirement.

- Exclusions: State exclusions directly and only when truly disqualifying. Do not repeat the same rule as both a requirement and a strict filter.

Examples of good requirements
Type | Example
--- | ---
Position and seniority | Current or recent experience as a Senior Backend Engineer.
Location | Based in London or Berlin.
Hard skill | Hands-on experience building production applications with Python.
Domain experience | Built payment processing systems for B2B products.
Leadership | Managed a team of at least five engineers.
Company and timing | Worked at Stripe before its Series B.
Skill in context | Used Python to build fraud detection systems.
Related skills | Backend development experience with Python and Go.
Exclusion | Must not currently work at Stripe.

## How filters work

Describe positions and locations in query or search_requirements. Pearch interprets the intent and maps the names for you.

Input | How Pearch interprets it
--- | ---
Positions | Maps equivalent names for the same role, including different job titles. Include seniority and specialization when they matter.
Locations | Interprets different names for the same geography and the intended geographic scope.

### Custom filters

- When to use custom_filters: custom_filters require Scale with Us and are not included in Free Trial or Self-Service. They can sharply reduce the number of matching candidates and exclude relevant people. Prefer search_requirements or query, which are available on all plans. In search_requirements, mark mandatory criteria with must_have: true. Contact [email protected] so we can help configure custom_filters for your search.

- AND and OR: In custom_filters, different fields combine with AND: every field must match. Alternative values in lists such as titles or locations combine with OR: any listed value can match.

- Exclusions with not_*: Use supported exclusion fields such as not_titles, not_locations, and not_companies to exclude matching candidates. A candidate matching any value in an exclusion list is excluded. The not_* prefix is supported only for the fields listed in the API reference.

- Location mistakes: Each location in custom_filters must be complete. For London OR Berlin, use two entries: “London, United Kingdom” and “Berlin, Germany.” Separate “London” and “United Kingdom” entries mean London OR anywhere in the UK. Choosing the wrong geographic level can also narrow the pool.

## Decide which candidates to show

Pearch scores candidates automatically. Choose whether your product shows only stronger matches or also offers lower-confidence results.

Score | Classification
--- | ---
4 | Great candidate
3 | Good candidate

- Preserve the API result order. Fast and Pro always score candidates. The public score is a coarse 1–4 grade, and Fast can rank candidates differently even when they all receive 4. Omit profile_scoring in new integrations. Leave fill_with_low_confidence_results set to false to keep low-confidence fallback disabled.

- Include lower-confidence results: Set fill_with_low_confidence_results to true to fill remaining slots after stronger candidates are exhausted. Keep them visibly distinct from Great and Good matches. Lower-confidence fallback results are free for search and scoring. Contact-reveal charges still apply when contact data is requested and found.

## Choose the search mode

Fast and Pro are two execution modes, not mandatory stages of the same search.

- Fast: Faster, with lower recall. A good fit for simple positions.

- Pro: Deeper analysis for complex positions. It can generally find more good candidates, but takes longer and costs more.

Increasing limit or time_budget does not make Fast equivalent to Pro. limit controls the number of results requested; time_budget controls how long the selected mode can run.

## Real-time profile data

Enable high_freshness to download the latest available profile directly from source websites in real time, including current roles and recently updated data.

- When to use high_freshness: Use high_freshness when your shortlist needs current data. This adds time and credits, so apply it where up-to-date information matters rather than to every candidate by default.

- Open to Work: There is no dedicated Open to Work filter. high_freshness may return is_opentowork: true when the source exposes it. A missing field is not proof that the candidate is not open to work.

## Filter contact availability or reveal contacts

Filtering and revealing are separate decisions. Availability filters constrain the candidate pool; reveal returns the contact data.

filter_out_no_emails, filter_out_no_phones, and filter_out_no_phones_or_emails require Scale with Us and are not included in Free Trial or Self-Service. These filters can sharply reduce the number of matching candidates. Contact [email protected] so we can help configure them for your search. This restriction also applies to their deprecated require_* names. Enabling a restricted filter without access returns HTTP 403 with code feature_not_enabled.

Parameter | Purpose
--- | ---
filter_out_no_emails | Require at least one personal email address.
filter_out_no_phones | Require at least one phone number.
filter_out_no_phones_or_emails | Require at least one phone number or personal email address.
reveal_emails | Return available personal email addresses in the search response.
reveal_phones | Return available phone numbers in the search response.

reveal_emails and reveal_phones are available on all plans. Use reveal only when the product needs contact data. Email and phone reveals are charged separately, only when the requested data is found.

## Load more results from the same search

Reuse thread_id when loading more results. Changing the criteria starts a new search without thread_id.

- When to offer Load more: Offer Load more when has_more_results is true. When false, do not request another page in the same thread. Do not infer availability from the result count or stop_reason alone.

- Request the next page using thread_id with limit and offset. Omit fixed search criteria or pass their saved values. To change those criteria, start a search with no thread_id. Page options include profile inclusion and exclusion lists, liked-profile exclusions, contact filters and reveals, freshness, insights, response formatting, and the time budget. Page filters apply before the offset. Request pages sequentially and deduplicate by a stable profile ID.

- Retry a page: Retry 429 and transient 5xx responses with backoff. Keep the same thread_id and pagination parameters when retrying a page; do not start a new search.

### Why the search stopped

Use stop_reason to explain why this response stopped. A short page alone does not mean there are no more candidates.

stop_reason | What it means | What to show next
--- | --- | ---
target_reached | The initial search reached its target, or the saved page filled and another matching candidate is available. | This does not mean the pool is exhausted.
time_budget_exhausted | The search or page filter reached its time budget while work remained. | Use has_more_results to decide whether to request another page.
candidate_pool_exhausted | Every fetched candidate was evaluated and no further matches remain in this thread. | Start a new search with broader requirements.

## Receive results when the search is ready

Use async: true for searches that may take longer than your product’s request timeout. The request returns immediately while the search continues.

- Receive a webhook: Include webhook_url with async: true. Pearch sends a POST to that URL when results are ready, with the same JSON response format as a synchronous search.

- Show progress: Keep the user informed while the search runs. Show the results when they arrive, then use thread_id for pagination.

## Capture explicit candidate feedback

The Search Feedback API lets customers mark a candidate as relevant or not relevant to a specific search. Pearch uses this feedback to improve quality for that search context.

Field | Send
--- | ---
chat_thread_id | The search thread_id.
search_call_uuid | The search response uuid.
linkedin_slug | The candidate’s LinkedIn slug.
Rating and reason | A positive or negative rating and a short reason when available.
author_type | person, agent, or mixed: who provided the feedback.

Capture explicit decisions rather than inferring relevance from clicks alone.

## Get help with your integration

Email [email protected] or message us in your shared Pearch group chat if you have one. We’ll help you work through questions and search issues.

For a search issue, include thread_id so we can reconstruct what happened. Send:

- thread_id and response uuid.

- What the user expected and what happened instead.

- Original user input and final API payload.

- Approximate timestamp.

- Example candidate slugs for false positives or false negatives.

Do not send only screenshots or rewritten summaries of the query.

## Understand additive credit pricing

The total cost combines the selected search mode, optional features, and successful contact reveals.

Your account’s prices may differ from the standard reference below. Read the current configuration through GET /v1/user; do not hard-code standard prices into billing logic.

Standard credit costs
Contact-availability filters and custom_filters, including custom-filter-only count requests, require Scale with Us and are not included in Free Trial or Self-Service. These filters can sharply reduce the number of matching candidates. Contact [email protected] so we can help configure them for your search. Search and count requests using query or search_requirements remain available on all plans.
Item | Standard credit cost
--- | ---
Fast search | 1 per profile
Pro search | 5 per profile
Profile insights | +1 per profile
high_freshness | +2 per profile
Contact-availability filter | +1 per profile
Email or phone reveal | +6 when found
Count with strict filters only | 2 per request
Count with query or search requirements | 5 per request
Count statistics | +5 per request

API REFERENCES
API documentation: https://apidocs.pearch.ai/
Official Python SDK: https://github.com/Pearch-ai/pearch-client
query: https://apidocs.pearch.ai/reference/post_v2-search#body-params-query
search_requirements: https://apidocs.pearch.ai/reference/post_v2-search#body-params-search_requirements
custom_filters: https://apidocs.pearch.ai/reference/post_v2-search#body-params-custom_filters
thread_id: https://apidocs.pearch.ai/reference/post_v2-search#body-params-thread_id
type: https://apidocs.pearch.ai/reference/post_v2-search#body-params-type
time_budget: https://apidocs.pearch.ai/reference/post_v2-search#body-params-time_budget
insights: https://apidocs.pearch.ai/reference/post_v2-search#body-params-insights
high_freshness: https://apidocs.pearch.ai/reference/post_v2-search#body-params-high_freshness
fill_with_low_confidence_results: https://apidocs.pearch.ai/reference/post_v2-search#body-params-fill_with_low_confidence_results
filter_out_no_emails: https://apidocs.pearch.ai/reference/post_v2-search#body-params-filter_out_no_emails
filter_out_no_phones: https://apidocs.pearch.ai/reference/post_v2-search#body-params-filter_out_no_phones
filter_out_no_phones_or_emails: https://apidocs.pearch.ai/reference/post_v2-search#body-params-filter_out_no_phones_or_emails
reveal_emails: https://apidocs.pearch.ai/reference/post_v2-search#body-params-reveal_emails
reveal_phones: https://apidocs.pearch.ai/reference/post_v2-search#body-params-reveal_phones
limit: https://apidocs.pearch.ai/reference/post_v2-search#body-params-limit
offset: https://apidocs.pearch.ai/reference/post_v2-search#body-params-offset
async: https://apidocs.pearch.ai/reference/post_v2-search#body-params-async
webhook_url: https://apidocs.pearch.ai/reference/post_v2-search#body-params-webhook_url
chat_thread_id: https://apidocs.pearch.ai/reference/post_v2-search-feedback#body-params-chat_thread_id
search_call_uuid: https://apidocs.pearch.ai/reference/post_v2-search-feedback#body-params-search_call_uuid
linkedin_slug: https://apidocs.pearch.ai/reference/post_v2-search-feedback#body-params-linkedin_slug
author_type: https://apidocs.pearch.ai/reference/post_v2-search-feedback#body-params-author_type
must_have: https://apidocs.pearch.ai/reference/post_v2-search#body-params-search_requirements
async: true: https://apidocs.pearch.ai/reference/post_v2-search#body-params-async
Search Feedback API: https://apidocs.pearch.ai/reference/post_v2-search-feedback
not_titles: https://apidocs.pearch.ai/reference/post_v2-search#body-params-custom_filters
not_locations: https://apidocs.pearch.ai/reference/post_v2-search#body-params-custom_filters
not_companies: https://apidocs.pearch.ai/reference/post_v2-search#body-params-custom_filters
[email protected]: mailto:[email protected]
POST /v2/search/submit: https://apidocs.pearch.ai/reference/post_v2-search-submit
GET /v2/search/status/{task_id}: https://github.com/Pearch-ai/pearch-client/blob/master/pearch-openapi.yaml
has_more_results: https://apidocs.pearch.ai/reference/post_v2-search#response-schemas
stop_reason: https://apidocs.pearch.ai/reference/post_v2-search#response-schemas
target_reached: https://apidocs.pearch.ai/reference/post_v2-search#response-schemas
time_budget_exhausted: https://apidocs.pearch.ai/reference/post_v2-search#response-schemas
candidate_pool_exhausted: https://apidocs.pearch.ai/reference/post_v2-search#response-schemas
Fast search: https://apidocs.pearch.ai/reference/post_v2-search#body-params-type
Pro search: https://apidocs.pearch.ai/reference/post_v2-search#body-params-type
Contact-availability filter: https://apidocs.pearch.ai/reference/post_v2-search#body-params-filter_out_no_phones_or_emails
Email or phone reveal: https://apidocs.pearch.ai/reference/post_v2-search#body-params-reveal_emails
Count with strict filters only: https://apidocs.pearch.ai/reference/post_v2-search-count
Count with query or search requirements: https://apidocs.pearch.ai/reference/post_v2-search-count
Count statistics: https://apidocs.pearch.ai/reference/post_v2-search-count
GET /v1/user: https://apidocs.pearch.ai/reference/get_v1-user

Integration guides: https://apidocs.pearch.ai/docs/product-integration
API reference: https://apidocs.pearch.ai/reference
Official Python SDK: https://github.com/Pearch-ai/pearch-client

MCP integration

Integrate Pearch into the agent inside this product using the remote MCP server https://mcp.pearch.ai/mcp.
1. The developer creates a Pearch API key for the platform and stores it in backend secret configuration as PEARCH_API_KEY. Do not ask them to paste it into chat.
2. Inspect the existing stack. Add a server-side MCP client using Streamable HTTP. Send Authorization: Bearer with PEARCH_API_KEY in the request header, and expose the selected Pearch tools to the product agent. This product integration uses the platform’s API key. End users do not sign in to Pearch or provide a key. Keep the key out of browser code, model context and output.
3. Discover tool schemas and call get_user_info for a free account check. Report account identity, credits and prices only to the platform administrator. Do not print full responses that contain credentials.
4. Prepare the first search with search_requirements. Let the user review and edit must-haves, preferences, and exclusions first. Preserve linked conditions within one requirement. Use the discovered schema for the exact argument structure. Run a small search only after the developer approves the use of Pearch credits. For this first test search, do not reveal contact data or refresh profiles unless requested.
5. Apply the recommendations to both backend behavior and the user interface. Flag any deviations in the existing product to the developer, with the impact and a suggested fix. Confirm unresolved product decisions before implementation.

search_people supports search_requirements. Use only arguments supported by the discovered tool schemas. If a selected feature, such as custom_filters or async with webhook_url, is not exposed by MCP, flag it to the developer and ask how to proceed. Do not invent arguments or silently drop the feature.

PRODUCT RECOMMENDATIONS

## Choose how users provide search intent

Input quality has the strongest influence on search quality. Good input is specific, structured, observable on a professional profile, and free of contradictions.

The hiring manager is the source of truth. Before searching, they should approve the target profile and business context, 2–6 real must-haves, nice-to-haves, and exclusions. A recruiter, integration, or LLM can structure the request, but should not silently choose essential criteria or trade-offs.

- Requirements first (recommended): Users enter structured search_requirements and review them before searching. They can see each criterion, mark must-haves and nice-to-haves, and resolve conflicts before spending credits. This gives them the most control.

- Query, review, then rerun: Users start with a natural-language query. Pearch converts it into search_requirements. Show those requirements for human review, then rerun the search with the approved criteria. This balances a simple input experience with explicit control.

- Query only: The search runs without a review of the interpreted requirements. This is supported but offers less control: vague, unstructured, or contradictory queries can lose important distinctions.

## Write clear, verifiable requirements

Each requirement should express one clear idea and be verifiable from a professional profile.

Avoid requirements that cannot normally be verified from a professional profile, such as willingness to accept a salary, availability to start on a specific date, or being a “team player.”

- Must-haves and preferences: Use 2–6 must-haves. A criterion is mandatory only if the customer would reject a candidate without it. Keep preferences as nice-to-haves so they improve ranking without excluding relevant candidates.

- Connected context: Keep a skill and how it was used together, or a company and the required period. These connected conditions belong in the same requirement.

- Exclusions: State exclusions directly and only when truly disqualifying. Do not repeat the same rule as both a requirement and a strict filter.

Examples of good requirements
Type | Example
--- | ---
Position and seniority | Current or recent experience as a Senior Backend Engineer.
Location | Based in London or Berlin.
Hard skill | Hands-on experience building production applications with Python.
Domain experience | Built payment processing systems for B2B products.
Leadership | Managed a team of at least five engineers.
Company and timing | Worked at Stripe before its Series B.
Skill in context | Used Python to build fraud detection systems.
Related skills | Backend development experience with Python and Go.
Exclusion | Must not currently work at Stripe.

## How filters work

Describe positions and locations in query or search_requirements. Pearch interprets the intent and maps the names for you.

Input | How Pearch interprets it
--- | ---
Positions | Maps equivalent names for the same role, including different job titles. Include seniority and specialization when they matter.
Locations | Interprets different names for the same geography and the intended geographic scope.

### Custom filters

- When to use custom_filters: custom_filters require Scale with Us and are not included in Free Trial or Self-Service. They can sharply reduce the number of matching candidates and exclude relevant people. Prefer search_requirements or query, which are available on all plans. In search_requirements, mark mandatory criteria with must_have: true. Contact [email protected] so we can help configure custom_filters for your search.

- AND and OR: In custom_filters, different fields combine with AND: every field must match. Alternative values in lists such as titles or locations combine with OR: any listed value can match.

- Exclusions with not_*: Use supported exclusion fields such as not_titles, not_locations, and not_companies to exclude matching candidates. A candidate matching any value in an exclusion list is excluded. The not_* prefix is supported only for the fields listed in the API reference.

- Location mistakes: Each location in custom_filters must be complete. For London OR Berlin, use two entries: “London, United Kingdom” and “Berlin, Germany.” Separate “London” and “United Kingdom” entries mean London OR anywhere in the UK. Choosing the wrong geographic level can also narrow the pool.

## Decide which candidates to show

Pearch scores candidates automatically. Choose whether your product shows only stronger matches or also offers lower-confidence results.

Score | Classification
--- | ---
4 | Great candidate
3 | Good candidate

- Preserve the API result order. Fast and Pro always score candidates. The public score is a coarse 1–4 grade, and Fast can rank candidates differently even when they all receive 4. Omit profile_scoring in new integrations. Leave fill_with_low_confidence_results set to false to keep low-confidence fallback disabled.

- Include lower-confidence results: Set fill_with_low_confidence_results to true to fill remaining slots after stronger candidates are exhausted. Keep them visibly distinct from Great and Good matches. Lower-confidence fallback results are free for search and scoring. Contact-reveal charges still apply when contact data is requested and found.

## Choose the search mode

Fast and Pro are two execution modes, not mandatory stages of the same search.

- Fast: Faster, with lower recall. A good fit for simple positions.

- Pro: Deeper analysis for complex positions. It can generally find more good candidates, but takes longer and costs more.

Increasing limit or time_budget does not make Fast equivalent to Pro. limit controls the number of results requested; time_budget controls how long the selected mode can run.

## Real-time profile data

Enable high_freshness to download the latest available profile directly from source websites in real time, including current roles and recently updated data.

- When to use high_freshness: Use high_freshness when your shortlist needs current data. This adds time and credits, so apply it where up-to-date information matters rather than to every candidate by default.

- Open to Work: There is no dedicated Open to Work filter. high_freshness may return is_opentowork: true when the source exposes it. A missing field is not proof that the candidate is not open to work.

## Filter contact availability or reveal contacts

Filtering and revealing are separate decisions. Availability filters constrain the candidate pool; reveal returns the contact data.

filter_out_no_emails, filter_out_no_phones, and filter_out_no_phones_or_emails require Scale with Us and are not included in Free Trial or Self-Service. These filters can sharply reduce the number of matching candidates. Contact [email protected] so we can help configure them for your search. This restriction also applies to their deprecated require_* names. Enabling a restricted filter without access returns HTTP 403 with code feature_not_enabled.

Parameter | Purpose
--- | ---
filter_out_no_emails | Require at least one personal email address.
filter_out_no_phones | Require at least one phone number.
filter_out_no_phones_or_emails | Require at least one phone number or personal email address.
reveal_emails | Return available personal email addresses in the search response.
reveal_phones | Return available phone numbers in the search response.

reveal_emails and reveal_phones are available on all plans. Use reveal only when the product needs contact data. Email and phone reveals are charged separately, only when the requested data is found.

## Load more results from the same search

Reuse thread_id when loading more results. Changing the criteria starts a new search without thread_id.

- When to offer Load more: Offer Load more when has_more_results is true. When false, do not request another page in the same thread. Do not infer availability from the result count or stop_reason alone.

- Request the next page using thread_id with limit and offset. Omit fixed search criteria or pass their saved values. To change those criteria, start a search with no thread_id. Page options include profile inclusion and exclusion lists, liked-profile exclusions, contact filters and reveals, freshness, insights, response formatting, and the time budget. Page filters apply before the offset. Request pages sequentially and deduplicate by a stable profile ID.

- Retry a page: Retry 429 and transient 5xx responses with backoff. Keep the same thread_id and pagination parameters when retrying a page; do not start a new search.

### Why the search stopped

Use stop_reason to explain why this response stopped. A short page alone does not mean there are no more candidates.

stop_reason | What it means | What to show next
--- | --- | ---
target_reached | The initial search reached its target, or the saved page filled and another matching candidate is available. | This does not mean the pool is exhausted.
time_budget_exhausted | The search or page filter reached its time budget while work remained. | Use has_more_results to decide whether to request another page.
candidate_pool_exhausted | Every fetched candidate was evaluated and no further matches remain in this thread. | Start a new search with broader requirements.

## Receive results when the search is ready

Use async: true for searches that may take longer than your product’s request timeout. The request returns immediately while the search continues.

- Receive a webhook: Include webhook_url with async: true. Pearch sends a POST to that URL when results are ready, with the same JSON response format as a synchronous search.

- Show progress: Keep the user informed while the search runs. Show the results when they arrive, then use thread_id for pagination.

## Capture explicit candidate feedback

The Search Feedback API lets customers mark a candidate as relevant or not relevant to a specific search. Pearch uses this feedback to improve quality for that search context.

Field | Send
--- | ---
chat_thread_id | The search thread_id.
search_call_uuid | The search response uuid.
linkedin_slug | The candidate’s LinkedIn slug.
Rating and reason | A positive or negative rating and a short reason when available.
author_type | person, agent, or mixed: who provided the feedback.

Capture explicit decisions rather than inferring relevance from clicks alone.

## Get help with your integration

Email [email protected] or message us in your shared Pearch group chat if you have one. We’ll help you work through questions and search issues.

For a search issue, include thread_id so we can reconstruct what happened. Send:

- thread_id and response uuid.

- What the user expected and what happened instead.

- Original user input and final API payload.

- Approximate timestamp.

- Example candidate slugs for false positives or false negatives.

Do not send only screenshots or rewritten summaries of the query.

## Understand additive credit pricing

The total cost combines the selected search mode, optional features, and successful contact reveals.

Your account’s prices may differ from the standard reference below. Read the current configuration through GET /v1/user; do not hard-code standard prices into billing logic.

Standard credit costs
Contact-availability filters and custom_filters, including custom-filter-only count requests, require Scale with Us and are not included in Free Trial or Self-Service. These filters can sharply reduce the number of matching candidates. Contact [email protected] so we can help configure them for your search. Search and count requests using query or search_requirements remain available on all plans.
Item | Standard credit cost
--- | ---
Fast search | 1 per profile
Pro search | 5 per profile
Profile insights | +1 per profile
high_freshness | +2 per profile
Contact-availability filter | +1 per profile
Email or phone reveal | +6 when found
Count with strict filters only | 2 per request
Count with query or search requirements | 5 per request
Count statistics | +5 per request
Reference: https://github.com/Pearch-ai/mcp_pearch

Integration guides: https://apidocs.pearch.ai/docs/product-integration
API reference: https://apidocs.pearch.ai/reference
Official Python SDK: https://github.com/Pearch-ai/pearch-client

Did this page help you?