This is an early release preview. You may encounter bugs.
Treg logo
Unclaimed

Tool connectors security

Treg

OpenRouter for agent tools. Join community here: https://discord.gg/6mQYYfFMAn

B 71/100 GitHub score ? This grade is derived from GitHub signals, not user votes. Open for the full breakdown.
No votes yet

01 / About

What Treg is.

Treg is a registry and proxy that gives an agent one base URL and one token for a catalogue of third-party tool endpoints. It lists 2,896 catalogued endpoints across 60 providers covering search-engine optimisation and backlinks, social and trends data, people and company enrichment, advertising, and scraping, billed per call from a cent with no account at the provider. Search is by the job rather than the vendor: treg catalog search "find a work email" returns matching endpoints with their prices, treg catalog get shows parameters and an example response, and treg call runs it.

A catalogued call resolves its credential in a fixed order: a tool your team registered for that provider, then a secret your team stored for it, then Treg's own key billed against the team's prepaid balance, so a key you already pay for takes precedence and those calls are not metered. An endpoint with no published price is refused rather than served free. Running out of balance returns HTTP 402 carrying balancemicro, estimatedcostmicro, and a topupurl so an agent can act on it without parsing prose.

Alongside the catalogue, a team registers its own tools. An endpoint is an upstream base URL plus credential bindings that inject secrets server-side; a CLI is a vendor binary such as stripe, gh, or vercel run with the credential injected, either locally or on the registry server; a skill bundles a SKILL.md recipe with its secrets and tools. treg scan previews what a project would share and treg upload registers it, reading the .env against about 80 known providers plus skill directories and installed catalogue CLIs. Calls can be made by prefixing the real upstream request with the proxy URL, which resolves the tool by host, injects the credential, strips the caller's token, and relays the rest unchanged. treg shell start opens a subshell where every registered CLI injects automatically, and treg calls and treg runs are the audit logs.

Agents reach the same registry over Model Context Protocol as well as the CLI. A Claude Code plugin installs with no token and walks the agent through the CLI, sign-in, and treg mcp install on first run; the same skill ships through the MiniMax plugin marketplace. A separate curated surface for the Claude connectors directory exposes only catalogue endpoints and separates read calls from write calls. The service is hosted at treg.to and can also be self-hosted.

Features

  • One token, many providers: a single credential reaches every catalogued endpoint, with no per-provider signup or scattered key files
  • Task-first search: the catalogue is grouped by what endpoints do, and competing providers for one capability are shown side by side with prices
  • Per-call pricing: catalogued calls are metered against a prepaid team balance, with $1.00 free on a new team and an HTTP 402 carrying machine-readable top-up details
  • Credential ladder: a team's own tool or stored secret always wins over Treg's key, and those calls are never metered
  • Faithful relay: the proxy relays rather than models the upstream and injects auth server-side, so callers never hold keys
  • CLI execution: treg run executes a vendor binary with the org credential injected, locally or on the registry server, and treg shell start injects for a whole session
  • Skill bundles: a SKILL.md recipe, its secrets, and its tools register together and can be shared across a team
  • Automatic registration: treg scan and treg upload detect .env keys, skill directories, and installed catalogue CLIs
  • MCP surfaces: a general endpoint for catalogue, team-owned, and imported tools, plus a curated connector surface that separates read from write calls
  • Audit logs: treg calls and treg runs record proxied requests and CLI executions

02 / Discussion CREDIBILITY-GATED

Discussion

Reading is open to everyone. Posting and voting need a verified identity or a GitHub grade of B or higher.

  • No discussions yet.

03 / Build

Build with Treg.

Browse the catalogue for frameworks, tools, and harnesses, each scored on real GitHub credibility.

Get Treg →

Browse the catalogue