Skip to main content

Agent Tooling

Dwellir builds tools and infrastructure that AI coding agents can use directly. This page covers the CLI, agent skills, migration automation, and the documentation endpoints that make Dwellir's docs consumable by agents without HTML parsing.

Dwellir CLI#

The Dwellir CLI gives agents full access to the Dwellir platform from the command line. Every command supports --json output wrapped in a standard {ok, data, meta} envelope, and the CLI auto-selects JSON when stdout is not a terminal.

Key capabilities for agents:

  • Endpoint discovery: dwellir endpoints list --json returns the full catalog of 150+ blockchain endpoints with connection URLs, node types, and ecosystems.
  • API key management: Create, rotate, and delete API keys programmatically with dwellir keys create --name "agent-key" --json.
  • Usage analytics: Query request counts, RPS, costs, and error logs with structured JSON output.
  • Built-in docs: dwellir docs list and dwellir docs get <topic> fetch Dwellir documentation as markdown directly in the terminal.

Install the CLI:

curl -fsSL https://raw.githubusercontent.com/dwellir-public/cli/main/scripts/install.sh | sh

See the full CLI documentation for all commands and flags.

Hyperliquid Agent Skill#

The Hyperliquid agent skill gives AI coding agents procedural knowledge for building on Hyperliquid through Dwellir's infrastructure. It follows the open Agent Skills standard, making it portable across 40+ AI coding agents including Claude Code, Cursor, and Windsurf.

What the skill provides#

Once installed, the skill automatically activates when an agent encounters Hyperliquid-related tasks. It equips the agent with knowledge of:

  • HyperEVM JSON-RPC: Query EVM state, deploy Solidity contracts (Chain ID: 999, gas token: HYPE)
  • Info API: Access market data, prices, order books, candles, funding rates, user positions and balances
  • gRPC streaming: Stream real-time L1 block data, fill executions, and order book snapshots
  • Order book WebSocket: Real-time L2 and L4 depth data from edge servers in Singapore and Tokyo
  • Native API routing: The skill teaches agents to route read operations through Dwellir endpoints and write operations (orders, transfers) through Hyperliquid's native API with EIP-712 signatures

Install#

npx skills add dwellir-public/hyperliquid-skills

This installs the skill into your project's .claude/skills/ directory. No manual invocation is needed.

RPC Migration Prompt#

The migration prompt automates switching a project's blockchain RPC endpoints from other providers to Dwellir. Copy it into your AI agent's context or use it as a system prompt. It walks agents through a structured 5-phase process: environment discovery, codebase scanning, compatibility matching, migration, and summary reporting.

View and copy the migration prompt
Migrate this project's blockchain RPC endpoints to Dwellir. Follow the phases below in order. Use subagents and background processes to parallelize work wherever your tooling supports it.

## Phase 1 — Environment & Endpoint Discovery

1. Check whether the Dwellir CLI is installed by running `dwellir --version`. If the command is not found, suggest the user install it:
   curl -fsSL https://raw.githubusercontent.com/dwellir-public/cli/main/scripts/install.sh | sh
2. Obtain the full list of Dwellir-supported chains, networks, and node types (full vs archive). Try one of these approaches in order until one succeeds:
   a. **CLI** (preferred): Run `dwellir endpoints list` to get the complete endpoint catalog.
   b. **Documentation**: If the CLI is unavailable or the user declines to install it, fetch https://www.dwellir.com/docs.md or https://www.dwellir.com/networks.md for the supported endpoint list.
   c. **Dashboard export**: As a last resort, ask the user to go to dashboard.dwellir.com/endpoints and press the Export button (top-left) to export all endpoints as CSV, Markdown, or JSON, then share the file with you.
3. Check whether the project uses separate configurations per environment (production, staging, development, etc.). If it does, ask the user to provide a Dwellir API key for each environment. If there is only one environment, ask for a single key.

## Phase 2 — Codebase Discovery

Scan the entire codebase in parallel where possible:

1. Find every RPC endpoint URL (look for domains like infura.io, alchemy.com, quicknode.com, chainstack.com, ankr.com, blast.io, drpc.org, and any other known RPC providers, as well as raw IP/port patterns and chain-specific gateway URLs).
2. Identify each endpoint's chain, network, and authentication method (API key in URL path, header, query param, or none).
3. Determine whether the code requires an archive node or a full node for each endpoint (look for calls to historical state such as eth_getBalance at old block heights, debug_*/trace_* namespaces, or large block-range log filters).
4. Check if the codebase interacts with Hyperliquid. If it does, suggest that the user install Dwellir's Hyperliquid Skills: npx skills add https://github.com/dwellir-public/hyperliquid-skills

## Phase 3 — Compatibility Matching

For each discovered endpoint:

1. Compare the chain + network against the Dwellir endpoints list from Phase 1. Note that some providers (especially for EVM chains) use chain ID-based naming in their URLs rather than chain + network names — resolve any ambiguity by calling the endpoint's RPC method for chain ID (e.g., eth_chainId) and comparing the result against the chain ID returned by the corresponding Dwellir endpoint to confirm they serve the same network.
2. If the code requires an archive node and Dwellir only offers a full node for that chain, mark the endpoint as unsupported and do NOT migrate it.
3. For EVM chains, check whether the codebase depends on client-specific response shapes (e.g., Geth/Erigon trace formats vs Reth, differences in debug_traceTransaction output, or Parity-style trace_* responses). Use web search if needed to understand current client-level differences. Flag any potential incompatibilities.

## Phase 4 — Migration

1. Create a new branch (e.g., chore/migrate-to-dwellir) — NEVER commit directly to main.
2. For each supported endpoint, replace the provider URL with the equivalent Dwellir endpoint URL and update the authentication to use the correct Dwellir API key for each environment. Preserve the existing configuration pattern (env var, config file, etc.).
3. If any endpoints, chains, or networks in the codebase are NOT supported by Dwellir, do not touch them.

## Phase 5 — Summary

Present a clear summary with:

- Migrated: list of endpoints successfully switched to Dwellir (chain, network, full/archive).
- Flagged: any EVM client compatibility concerns the user should verify.
- Not supported: list of endpoints/chains/networks Dwellir does not currently support, along with whether each requires a full or archive node and the estimated monthly request volume if determinable from the code. Ask the user to reach out to support@dwellir.com or the team on https://t.me/dwellir with this list so Dwellir can evaluate adding support.

If there are any questions about supported RPC methods or Dwellir services, consult https://www.dwellir.com/docs/llms.txt and https://www.dwellir.com/docs.md for authoritative reference.

Commit the changes and ask the user whether you should push the branch to origin and open a pull request.

Documentation for Agents#

Dwellir's documentation is built with AI agent ergonomics in mind. Every page is available as clean markdown, and multiple discovery mechanisms help agents find the right content without HTML parsing.

llms.txt#

The site follows the llmstxt.org specification. Agents can fetch a structured index of all available content:

EndpointDescription
/llms.txtSite-wide index with links to all networks, blog posts, and documentation
/docs/llms.txtCurated documentation index with network guides and key docs
/docs/llms-full.txtComplete documentation concatenated into a single file for large-context agents

Markdown endpoints#

Append .md to any documentation URL to get the content as plain markdown with Content-Type: text/markdown headers:

https://www.dwellir.com/docs/ethereum.md
https://www.dwellir.com/networks/ethereum.md
https://www.dwellir.com/blog/hyperliquid-rpc-providers-2025.md

These endpoints strip all JSX/MDX components, rewrite internal links to absolute URLs, and include YAML frontmatter with title and description metadata.

Smart 404 responses#

When a documentation page is not found, the API returns fuzzy-matched suggestions instead of a generic error. Agents receive:

  • A list of similar pages ranked by confidence score
  • Direct links to the matching markdown URLs
  • Fallback links to /docs/llms.txt and /llms.txt for broader discovery

Single exact matches (confidence 1.0) trigger an automatic redirect. This means minor typos or case mismatches resolve automatically.

Content negotiation#

All markdown endpoints set response headers that agents can use for content negotiation:

  • Content-Type: text/markdown; charset=utf-8
  • Link header with rel="canonical" (HTML version) and rel="alternate" (markdown version)
  • X-Robots-Tag: noindex, follow to prevent search engine indexing of markdown versions

Index endpoints#

Two index endpoints provide structured lists of all available content:

  • /docs.md - All documentation pages with titles, descriptions, and markdown links
  • /networks.md - All supported networks with availability, regions, and pricing summaries

Next steps#