Every way TachiBot can make money · researched mid-2026
Each avenue: how it works, real 2026 numbers, effort, and realistic revenue at the current audience size. Verdicts: now funnel later skip
Sell scarce orchestration expertise as fixed-price packages — the only avenue that pays this quarter.
"MCP orchestration setup + 3 custom workflows/skills" at $500–$2K fixed, repeatable. Market rates: AI consulting $150–$500/hr; MCP server builds $25K–$50K (SMB MVP) to $60K–$120K (production); a single custom agent ~$30K. You don't need enterprise deals — you need a landing-page offer and one case study. Constraint is sales, not capability.
Prereqs: offer on landing page · 1 case study · calendly link
Keep the engine open; gate convenience and team features behind a Polar.sh subscription.
Credible gates: usage-analytics dashboard (extend usage_stats), managed key vault / config sync, team profiles, curated premium workflow packs. Billing rail: Polar.sh (5% fee, GitHub-native, subs + license keys + issue bounties, ~10-min setup). Sustainable OSS devs at $1K+/mo combine 2+ streams. Aspirational ceiling: 21st.dev Magic hit $10K MRR in six weeks — but with proprietary generation, not a BYOK router.
Prereqs: the gated features must exist — analytics + key sync are the wanted ones
Hold the keys, meter usage, take margin — the only durable-margin model.
OpenRouter adds no token markup (~5.5% card fee, 5% BYOK fee above 1M req/mo) — your margin is what you layer on top, typically 10–30%. Realistic MCP SaaS outcomes: $500–$3K/mo for most that earn real money; $3K–$10K+ for top creators. Removes the BYOK friction that caps every other avenue. Heavy for a solo dev: key custody, metering, abuse/cost risk, support. Build the skeleton in 90 days; launch when the funnel fills.
Prereqs: metering loop · audience · Polar sub wiring
Distribution, not revenue — list everywhere, expect $0 direct.
Anthropic Connectors Directory: self-serve, review-gated, no paid tier, no rev-share (users pay Anthropic for Claude plans, not you). Smithery ~$30/mo hosting, returns $0; Glama keeps sub revenue; none do rev-share; all reward freemium with placement. Platforms that DO pay — MCPize (80–85% rev-share) and Apify (80% after compute, $4M+ paid to devs) — only work for hosted/metered servers, which BYOK TachiBot is not.
Prereqs: none — submit and forget
Turn on now — free trust signal; moves the needle only after the audience does.
GitHub Sponsors takes 0%; Polar 5% but adds products/bounties — most run both. Reality check: <12% of OSS devs earn anything; typical Sponsors ramp $80→$410/mo; $1K+/mo requires a second stream. Docs quality is the biggest conversion driver. Polar issue-bounties let users fund specific features — a nice fit for a tool roadmap.
Prereqs: FUNDING.yml · Polar account · good docs
Teach the jury/council protocols free; monetize indirectly. Don't sell a course.
TachiBot's differentiators (jury/council synthesis, 31 prompt techniques, ToT/reflexion skills, zero-cost local jurors) are teachable and under-covered. But mid-tier paid newsletters need ~47K readers for ~$6.4K/mo, and courses are in structural decline (<5% completion; AI erodes static content). Publish 2–3 strong free pieces that feed Sponsors, Pro tier and services.
Prereqs: 2–3 pieces on jury/council patterns · cross-post to landing
Charge agents per-call in USDC — strategically interesting, commercially early.
The 2026 stack (Coinbase/Cloudflare x402, Stripe Machine Payments, Google agentic payments) enables account-less per-call pricing, and Anthropic is a core x402 Foundation member. But volume is ~$28K/day and largely synthetic despite a ~$7B ecosystem valuation. The one place BYOK stops being a curse — hosted TachiBot tools as x402-priced endpoints. Wire into the hosted tier as a forward option; don't build on it.
Prereqs: hosted tier first — this rides on avenue 3
Thin and unconfirmed — never plan around it.
Could not confirm a paying OpenRouter affiliate program in 2026 — only its fee structure surfaced. The 1%-of-referred-spend program found belongs to LLMGateway, a competitor. If a program opens, adding referral links is trivial; until then this is a rounding error.
Prereqs: n/a — opportunistic only
Keep AGPL-3.0. Relicensing solves a problem you don't have yet — in either direction.
Current license is AGPL-3.0 — already the strongest hosted-moat position: anyone offering TachiBot as a network service must open their source. As sole copyright holder you can dual-license commercially today, or relax to MIT later; MIT now is a one-way door that gives away hosting rights. BSL/FSL (Sentry-style) exist to stop free-riders cloning your success — at 15 stars there's no success to free-ride. If an enterprise's AGPL ban ever blocks a deal, that's a commercial-license sale, not a reason to switch.
Prereqs: keep sole copyright (CLA note in CONTRIBUTING) to preserve dual-license rights