Business judgment on AI products
One local doorway for every AI assistant to reach its tools, with secrets locked on the machine instead of pasted everywhere.
01
Start inside the user's day · Public facts + workflow reasoning · 2026-08-28
One local doorway for every AI assistant to reach its tools, with secrets locked on the machine instead of pasted everywhere.
Public materials do not yet show how users complete this job today or what they replace.
The product targets friction in this job, but public user evidence does not yet show the cost, frequency, or consequence of leaving it unsolved.
Each extra tool an AI uses gets reconfigured in every app, often with keys in plain text. Don't build another tool bazaar—sell “configure once, use everywhere, keys never leave the machine” to teams running several assistants at once.
It promises a simpler way to complete this job: One local doorway for every AI assistant to reach its tools, with secrets locked on the machine instead of pasted everywhere. The exact adoption motive and repeat use are not yet verified.
① Whether paying Teams customers appear (zero disclosed so far) — the only revenue; path; ② Whether native MCP lazy loading ships across the three big clients (Claude Code /; Codex / Cursor) — when it does, this is over; ③ GitHub stars and forks three months out — a measure of adoption as a referenc…
Keep watching. It promises a simpler way to complete this job: One local doorway for every AI assistant to reach its tools, with secrets locked on the machine instead of pasted everywhere. The exact adoption motive and repeat use are not yet verified.
find the fixed cost users burn on every request (tokens, config, time), quantify it into a measurable promise ("save 74-91%"), and publish the measurement method — efficiency tools persuade with reproducible data. free open source for reputation, team subscription for governance (shared config, per-member local keys) — the smallest viable business loop for a solo builder, even if the future is doubtful.
Open-source core (MIT, local app free) plus a Toolport for Teams subscription: free up; to 5 people, then $39/month (for up to 5) plus $12/person/month, identical hosted or; self-hosted, ~$390/year annual. Paying customers: none disclosed (… ① Whether paying Teams customers appear (zero disclosed so far) — the only revenue; path; ② Whether native MCP lazy loading ships across the three big clients (Claude Code /; Codex / Cursor) — when it does, this is over; ③ GitHub stars and forks three months out — a measure of adoption as a referenc…
One local doorway for every AI assistant to reach its tools, with secrets locked on the machine instead of pasted everywhere.
It promises a simpler way to complete this job: One local doorway for every AI assistant to reach its tools, with secrets locked on the machine instead of pasted everywhere. The exact adoption motive and repeat use are not yet verified.
An English validation note will follow from the public evidence.
The product claims to help users complete: “One local doorway for every AI assistant to reach its tools, with secrets locked on the machine inst”. User evidence has not yet verified pain intensity or the cost of doing without it.
The assessment is recorded; an English explanation is pending.
The assessment is recorded; an English explanation is pending.
The assessment is recorded; an English explanation is pending.
02
Market comparison
The Chinese–English market comparison is not complete yet. A conclusion follows only after its coverage and verifiable evidence are recorded.
03
The call and next move come first; the full read retains the evidence and counterevidence.
Configure every MCP server once, share them across all your AI clients through one local gateway — tools load on demand to save tokens, secrets stay in the OS keychain.
Independent developer "Tyler" (GitHub tsouth89, the launch platform @tsouth2), a one-person project. First launched 2026-06-23 under the name "Conduit"; the third launch in August 2026 hit daily rank #5 with 149 points, version v1.12.0. No funding, relying on GitHub Sponsors. Site: toolport.app.
Read: this is the most direct productization of "configuration complexity silently kills adoption." Going from zero to v1.12 in two months, with a reproducible benchmark, signed cross-platform installers, and near-daily commits — the execution puts it at the top of its size class.
Two things. "Flat tool loading": every MCP server used to send its full tool list to the model on every request, so the fixed token overhead grew with every server you added. Meta-tool on-demand discovery replaces the once-per-request full dump.
"Configure every client separately": the same MCP config plus plaintext API keys used to be pasted into each AI client individually; changing one server meant syncing N files and always missing one. Toolport replaces multi-client config drift — set up once, available everywhere.
Open-source core (MIT, local app free) plus a Toolport for Teams subscription: free up to 5 people, then $39/month (for up to 5) plus $12/person/month, identical hosted or self-hosted, ~$390/year annual. Paying customers: none disclosed (third-party research found zero).
Read: free for individuals to build reputation, team subscription selling governance (shared config, per-member local keys) — the smallest viable business loop for a solo builder. But a $39/month team tier competing against free MCP gateways from Microsoft, Docker, IBM, and Kong is a thin lane.
| Dimension | Call |
|---|---|
| Founder-product fit | The founder's real identity is barely verifiable (just "Tyler"); the product has to prove the pain itself |
| Product insight | Turned a quantifiable waste — tool definitions eating context — into a headline feature backed by a reproducible measurement |
| Execution quality | Cross-platform installers, Rust gateway, keychain secrets, integrity fingerprints, near-daily commits. Solid |
| Timing | The core capability is being absorbed natively by clients (Claude Code already lazy-loads MCP schemas) — the window may be one or two release cycles |
Among this batch, the strongest technical execution and the weakest business proposition. Three things worth keeping:
First, turn quantifiable waste into a headline. "Every request wastes 24K tokens on tool definitions" is a number every developer gets instantly, and it ships with a public reproducible benchmark. Infrastructure tools persuade through measurement, not slogans. Anyone building efficiency tooling should copy this.
Second, the security design is a reference implementation for MCP gateways. Fingerprinting tool definitions (alert when a server changes a tool, guarding against poisoning), human approval for destructive calls, secrets only in the keychain — this whole set of defenses is portable to any product where agents call external tools.
Third, the structural risk is the fatal one. Lazy loading is being absorbed natively by the clients — Claude Code already defers MCP schemas — and once every client supports on-demand loading, the meta-tool differentiator is zero. Meanwhile Microsoft, Docker, IBM, and Kong all ship free MCP gateways, so a solo developer is asking for $39/month next to platform freebies. Its real value is closer to "the pioneer's best practices": it assembled everything an MCP gateway should have in one place. Treat it as a reference implementation to copy, not an investment.
① Whether paying Teams customers appear (zero disclosed so far) — the only revenue path ② Whether native MCP lazy loading ships across the three big clients (Claude Code / Codex / Cursor) — when it does, this is over ③ GitHub stars and forks three months out — a measure of adoption as a reference implementation
Product logic: find the fixed cost users burn on every request (tokens, config, time), quantify it into a measurable promise ("save 74-91%"), and publish the measurement method — efficiency tools persuade with reproducible data.
Security design: tool-definition hash fingerprints, human approval for destructive calls, secrets that only ever live in the keychain — a portable defense set for any "agent calls external tools" product.
Pricing structure: free open source for reputation, team subscription for governance (shared config, per-member local keys) — the smallest viable business loop for a solo builder, even if the future is doubtful.
Worth watching, but hard to bet on. The technology deserves copying; the business model invites doubt — when platforms build the free version into their clients, an independent gateway's window may be one or two release cycles. Check back in three months against paying customers and client-side native loading.
05
Use these searches when the official site is missing or the current link is only a lead.