01Describe the exact target
Include role, industry, geography, recency, evidence requirements, exclusions, and desired result count.
Research people, projects, and outreach evidence.
Firsay searches live sources, verifies role and project evidence, removes duplicates, and prepares results for review. It is a research assistant, not a guarantee that every public profile is current or complete.
Quick start
02Apply and review filters
Check parsed roles, industries, sources, and recency before spending provider credits.
03Run and verify
Open the canonical BIO and evidence links. Report incorrect results instead of treating the score as proof.
Core workflows
People on X
Find founders, BD, partnerships, community, marketing, and other operators with explicit BIO evidence.
People and communities on Telegram
Search accessible public entities and indexed sources. Telegram does not provide a complete global BIO index.
Early projects and posts
Find launches, betas, waitlists, hiring posts, and other time-sensitive signals with first-party links.
Outreach campaign
Generate review-first drafts from verified fields. Sending remains mock-only until a real sender is explicitly configured.
How to read results
Evidence beats score.
A high score means the available fields matched the rules. Always inspect the BIO fragment, project association, recent activity, and direct contact before outreach.
Fewer results can be correct.
When only 4 of 50 profiles satisfy strict BIO, project, category, recency, and contact requirements, Firsay returns 4 instead of padding the list with weak matches.
Costs and tokens
Research capacity
Larger targets and strict verification require more source checks. Firsay may return fewer results when the available evidence does not meet the requested rules.
LLM tokens
Deterministic X and web searches normally show 0 LLM tokens. Optional classification can use Gemini, OpenRouter, or another configured model.
FAQ
Why did Firsay return fewer leads than requested?
The requested count is a target, not permission to invent or weaken evidence. Check the rejection summary to see whether profiles lacked project, category, recency, or contact proof.
Why was the run expensive when LLM tokens were zero?
The cost came from discovery and verification providers. Canonical BIO hydration, recent-post checks, project followings, and web verification are separate paid requests.
Why is a contact missing?
Many people do not publish a Telegram handle, email, or open DM path. Firsay labels the field as missing instead of guessing private contact information.
Can Firsay send outreach automatically?
Campaigns support drafting, review, approval, queues, and a mock sender. Real sending needs platform-specific safety controls and explicit configuration.
Does Firsay search all of Telegram?
No. Telegram has no complete global public BIO index. Results depend on accessible public entities, configured sources, indexed messages, and canonical profile checks.
What should I include in a bug report?
Include the exact task, selected filters, expected behavior, actual behavior, run time, result count, and a screenshot. Never include API keys, passwords, session strings, or private invite links.