# 15 Qualified Leads a Day: How Nikolay Lechev's OpenClaw agent Ross prospects for his solo consultancy for about $50 a month

Nikolay Lechev runs Omniflow, a one-person consultancy that fixes onboarding and activation for SaaS companies. Prospecting by hand gave him maybe 2-3 qualified leads on a good day. He set up an OpenClaw agent named Ross on a VPS, and three sub-agents now search the web and YC batches, X and LinkedIn for SaaS founders while an orchestrator checks, deduplicates and tightens the search every day. He wakes up to about 15 qualified leads in Slack, each with the reason it fits, plus outreach and LinkedIn drafts that he reviews before anything goes out. API costs first ran past $1,500 a month. Model routing and caching brought them to about $30-50.

**Try this step:** <strong>One agent on a server, talked to in Slack.</strong> Ross is an OpenClaw agent on a Hostinger VPS using their Docker container. Slack is where his clients and notes already live, so that is where he talks to Ross.


| | |
|---|---|
| **Who** | Nikolay Lechev, founder of the onboarding consultancy Omniflow, who writes at nikolaylechev.com. LinkedIn and GitHub lechev, linked from his site, verified 2026-11-15 |
| **Problem** | No sales team and no ads; manual prospecting between client work found 2-3 qualified leads on a good day |
| **Stack** | OpenClaw on a Hostinger VPS (Docker); Anthropic models only (Haiku, Sonnet, Opus); Brave Search API; Slack; three lead-search sub-agents plus an orchestrator; cron jobs; a custom dashboard; RevenueCat and Bird skills |
| **Confidence** | Medium (first-person; leads per day stated, conversions not counted) |
| **Last verified** | 2026-11-15 |

## The problem
His pipeline was content, referrals and manual outreach, and outreach was the bottleneck. He spent hours each week hunting for the right companies, checking their signals and writing personal emails. "It worked, but it didn't scale," and it ate into client work. Most leads came from scrolling LinkedIn and X between projects.

## The blueprint (as he describes it)
1. **One agent on a server, talked to in Slack.** Ross is an OpenClaw agent on a Hostinger VPS using their Docker container. Slack is where his clients and notes already live, so that is where he talks to Ross.
2. **Three searchers and one checker.** Three sub-agents search for qualified SaaS founders: one on the web (Brave API) and YC batches, one on X, one on LinkedIn. An orchestrator checks the quality of their output, removes duplicates and improves the search criteria daily.
3. **A schedule, not a chat.** A morning report at 7 AM with leads, drafts and trends. Heartbeat checks every 4 hours that only scan and log. At 10 PM Ross picks 3-5 tasks from a markdown backlog, drafts outreach, content or research, and saves it to a review folder.
4. **Drafts, not sends.** Ross writes 2-3 LinkedIn post drafts a day, cold email copy and competitor analysis. "I review and edit everything before it goes out."
5. **Short context files.** He restarted twice after long soul, memory and user files made Ross hallucinate. He had them rewritten as concisely as possible: "Give them what they need to start, not everything you know."
6. **Cheap models first.** API costs kept climbing past $1,500 a month because every task hit Opus. Ross researched fixes overnight: Haiku for scraping and lookups, Sonnet for daily work, Opus only for real reasoning, prompt caching, rate limits and heartbeats that log without deep analysis.
7. **Every mistake becomes a rule.** Failures go into skill files, and Ross reviews which leads convert and adjusts his own search criteria.

## Stated results
- Before Ross: 2-3 qualified leads a day by hand. Now: "I get 15 qualified leads every morning, each with context on why they're a fit."
- "I've had multiple calls booked directly from leads Ross surfaced." He gives no number.
- API costs: from $1,500+ a month to roughly $30-50, "A 97% reduction." The headline puts the whole setup at about $50 a month.
- Setup took about a day, including two full restarts.

## Caveats
- The post is from Feb 20, 2026, one month into the setup. Nothing newer is cited here.
- No count of calls, clients or revenue from the leads.
- Ross drafts outreach and posts; he reviews and sends. The post doesn't say how many emails go out.
- Weak spots he lists: LinkedIn needs manual token refreshes, Brave API rate limits, hard-to-debug sub-agents, and a YC scraping step that broke when a service went down.
- He links Hostinger's OpenClaw hosting page; the post doesn't say whether that is a paid link.
- Handles: LinkedIn and GitHub (lechev) linked from his site. No X, Bluesky, Threads, Instagram or TikTok profile verified.

## FAQ
### How does Nikolay Lechev use an AI agent to find leads?
His OpenClaw agent Ross runs three sub-agents that search the web and YC batches, X and LinkedIn for SaaS founders. An orchestrator checks and deduplicates the results, and about 15 qualified leads arrive in Slack each morning with the reason each one fits.

### How much does it cost to run?
He says roughly $30-50 a month in API costs after cutting them from over $1,500, mostly by sending simple tasks to Haiku and caching prompts.

### Does the agent send outreach on its own?
No. Ross drafts outreach, LinkedIn posts and research, and he reviews and edits everything before it goes out.

### Has it brought in clients?
He says multiple calls were booked from leads Ross found, but he gives no number of calls or clients.

## Sources
- https://nikolaylechev.com/blog/how-i-built-a-50-month-ai-employee-that-finds-me-15-qualified-leads-a-day (primary: Nikolay Lechev's first-person post "How I built an AI employee", Feb 20, 2026)
- https://www.omniflow.so/ (his consultancy, linked from the post)

---
*Want agents like these working your distribution? [Get It Built](/get-it-built). Get new teardowns weekly: [subscribe](/subscribe) (double opt-in).*