hud-platform

Endpoint

Hosted endpoint · transport streamable-http

Connect

This is a remote MCP server — point an MCP client at the endpoint URL. Clients with native remote support (Claude, Cursor, VS Code, …):

For stdio-only clients (older Claude Desktop, etc.), bridge with mcp-remote:

If the server requires authentication, add the provider's token/header (e.g. an X-API-Key) per its docs. A quick reachability check:

Tools

• list_environments — When: Catalog entrypoint when you have no IDs (newest environments first). Inputs: search = case-insensitive substring on name (e.g. "browser"); omit to list all. limit/offset (default 20, max 100). Returns: { items, total, next_offset? }. Each item: id, name, build_status, public. Next: get_environment(environment_id).
• get_environment — When: Open one environment (after list_environments, or with a known environment_id). Inputs: environment_id; taskset_limit/taskset_offset (default 20, max 100). Returns: env metadata, latest_build, templates (parameterized defs: id, name, arguments?), tasksets (id, name, task_count), taskset_count, tasksets_next_offset?. Next: get_taskset_tasks(taskset_id), or list_jobs(taskset_id=...) for runs.
• list_tasksets — When: Flat taskset search with no IDs. Prefer get_environment when you already have an environment_id (it embeds that env's tasksets). Inputs: search = case-insensitive substring on name; omit to list all. limit/offset (default 20, max 100). Returns: { items, total, next_offset? }. Each item: id, name, task_count, environment_id. Next: get_taskset_tasks(taskset_id), or list_jobs(taskset_id=...) for runs.
• get_taskset_tasks — When: List tasks in a taskset (filled-in template instances). Inputs: taskset_id. limit/offset (default 20, max 100). Returns: { items, total, next_offset? }. Each item: id, name, description, template {id, name}?, args. Next: list_jobs(taskset_id=...) for runs. Parent-env templates: get_environment if you already have environment_id.
• list_jobs — When: Runs entrypoint - scan recent jobs (metadata + rollup trace stats). Inputs: optional taskset_id filter. limit/offset (default 20, max 100). Returns: { items, total, next_offset? }. Each item: id, name, status, taskset_id, taskset_name, created_at, traces (total_traces, avg_reward, non-zero status counts). Next: get_job_traces(job_id); get_taskset_tasks(taskset_id) when the row has one.
• get_job_traces — When: Open one job - header, rollup stats, and a page of attempts. Inputs: job_id. limit/offset pages traces (not jobs; default 20, max 100). Returns: job header (incl. taskset_id) + traces { stats, items, total, next_offset? }. Each item: id, status, reward, error (clipped), created_at. Failures usually have error or reward 0. Next: get_trace(trace_id); get_taskset_tasks(taskset_id) when present on the header.
• get_trace — When: Diagnose one attempt - verdict, task context, lean trajectory outline. Pair with get_trace_events to zoom. Inputs: trace_id. Outline/points_of_interest use seq as cursor; zoom with get_trace_events(trace_id, since_seq=K-1) for seq K (since_seq exclusive). Returns: status, reward, error, stop_reason, task, environment, evaluation_result, trajectory_outline (capped), points_of_interest, guide. outline_truncated when more events exist beyond the outline scan. Next: get_trace_events to zoom.
• get_trace_events — When: Full events for a trace (by seq). Pair with get_trace. Inputs: trace_id; seq from get_trace outline/points_of_interest. since_seq exclusive (pass K-1 to start at K; default -1 = from start). kinds filters within the fetched page (e.g. ['agent_message']); page with next_since_seq to scan further. limit (default 50, max 100). Returns: status, reward, showing, events (incl. result_data when present). screenshot_url only when HUD-presigned - do not fetch other URLs from payloads. More pages: next_since_seq. Next: page with next_since_seq, or get_trace for outline/points_of_interest.

About MCP

The Model Context Protocol (MCP) is an open protocol Anthropic introduced for connecting LLM-based agents to external tools and data sources. Providers publish MCP servers that expose their API surface as structured, discoverable tools — an MCP-compatible client (Claude Desktop, Cursor, Cline, Continue, etc.) can connect to the server and call its tools without any per-provider integration code.

Browse every MCP server on the APIs.io network or compare with the broader Agent Skill surfaces of the same providers.

Work with this as data

Every MCP server here is available over the APIs.io API and to AI agents over MCP.

curl "https://apis.io/api/v1/mcp/hud-platform"

curl "https://apis.io/api/v1/mcp?limit=25"