Skip to main content
The registry proxy connector lets an agent discover and call a large catalogue of third-party API endpoints (SEO, enrichment, social data, scraping, ads) through a single token — vendor credentials are injected server-side by the registry, so the agent never holds keys.
Bring your own account. PraisonAI ships no registry code and no default endpoint. You supply your own account token or a self-hosted base URL. The connector is disabled until TOOL_PROXY_URL is set.

Quick Start

1

Install and configure

2

Give the three functions to an agent

The agent searches the catalogue, describes a matching endpoint, then calls it — one token, no vendor keys in the agent.

How It Works

The connector mirrors PraisonAI’s deferred tool-search bridge with three functions over the registry’s HTTP surface.

The Three Functions

Each function is an agent tool that defaults to the TOOL_PROXY_URL / TOOL_PROXY_TOKEN environment variables. Free capability search returning catalogue matches for a natural-language query.

registry_describe

Free read returning an endpoint’s parameters, price-per-call and an example response.

registry_call

Paid invoke. The registry injects the upstream vendor credential server-side.

Spend Guards

registry_call accepts optional budget guards that deny and report a call as a tool-result error before any money is spent.
  • Per-call price is read from registry_describe, so guards work without vendor-specific configuration.
  • Guards fail closed: if a budget is set but the price cannot be validated, the call is denied.
  • max_session_spend persists across separate registry_call invocations in the same process, scoped per (proxy_url, token) so distinct accounts never share a budget.

Security Model

The connector is designed so a prompt-injected agent cannot leak the proxy token or spend uncontrolled money.
The registry injects the upstream vendor credential — the agent only ever holds the single proxy token, never vendor keys.
Auth, insufficient-balance, upstream and timeout failures are returned as {"error": ...} tool results, so a failed call never crashes the agent.
The ambient TOOL_PROXY_TOKEN is attached only when the URL is also env-derived. An agent-supplied proxy_url never receives it, preventing a prompt-injected agent from sending the token to an attacker-controlled endpoint.
registry_call registers at medium risk in the approval registry because it spends money; registry_search and registry_describe are free reads.

Zero-Config Skills Path

Registries that expose skill install <name> place skills in ./.claude/skills/, which PraisonAI already scans — so registry-installed skills are discoverable with no configuration.
The compatibility scan (plus an ancestor walk for monorepos) lives in core praisonaiagents/skills/discovery.py.

Environment Recipe

Use the registry’s authenticated CLI passthrough inside a sandbox or environment setup: so the agent’s shell commands get authenticated vendor CLIs with no keys in the container.
The agent then runs authenticated vendor CLIs through the registry (for example run gh -- pr list) without any vendor key in the container. Sandbox and board workers inherit the same via the environment definition.

Troubleshooting

If TOOL_PROXY_URL is unset, every call returns: Tool registry proxy is not configured. Set TOOL_PROXY_URL (and optionally TOOL_PROXY_TOKEN) to enable the connector.
Calls return httpx not installed. Install with: pip install 'praisonai-tools[registry-proxy]'. Install the extra to add httpx.
All failures come back as {"error": ...} tool results: authentication failed (HTTP 401/403), insufficient balance (HTTP 402), upstream 4xx/5xx passed through, and request timeouts.
Two compatibility items are tracked in PraisonAI-Tools #79 — auth-header configurability and price-field tolerance. Behaviour may broaden as these land.

Tools Overview

Browse PraisonAI tool documentation

Custom Tools

Build your own agent tools