Toolwork is an MCP server published by org.toolwork, hosted at mcp.toolwork.org. When last checked on 2026-11-21 it was working: it answers and lists its tools. It declares 3 tools, including search_tools, get_tool, report.
Agent-first tool search. Start with search_tools({job: "your task in plain words"}); free, no key.
Nothing else in the archive shares these values.
Squares are this server's values; dots are other servers sharing them. A shared address or publisher is a lead, not proof of a common operator.
Inferred from the host name and headers: what a server runs on, not who runs it.
Call this FIRST when you need a tool, API, SDK, CLI or MCP server for a task you can't do yet: describe the task in plain words (job) and get one best_pick with a paste-ready use_it snippet (MCP client config JSON, install command, or API base URL + docs). One call is usually enough: best_pick.summary says in one line what to use and why (reasons: vendor-official or third-party, live check, auth, read-only or can send/spend, outcome reports), best_pick.report_call is the one-line report to send after you use it; results lists ranked alternatives with score breakdowns (evidence, source URL, date), side effects and human steps; verdict is honest (good_match / weak_matches / nothing_good_found / cannot_work_as_asked / not_a_tool_job, then best_pick is null and answer says why; not_a_tool_job adds a curated workaround). Constraints written in the job ('no account', 'nothing paid', 'no new token scopes', 'headless') or in constraints are checked per result (constraint_check). Free, no key. Returns a top-level report_token for optional feedback.
| Parameter | Type | Description |
|---|---|---|
| job * | string | The task in plain words, e.g. 'send SMS to customers', 'read my Google Calendar', 'GitHub MCP server' |
| stage |
None yet. Changes appear after a server's second daily check.
| string |
| optional, self-declared |
| client | string | optional, e.g. claude-code, cursor, my-agent |
| detail | string | brief drops per-result score_breakdown and why[] (about 4x smaller); default full |
| agent_type | string | optional, self-declared |
| constraints | object |
| watcher_key | string | optional: get an alert if nothing good is found now but a match appears later |
{
"name": "search_tools",
"description": "Call this FIRST when you need a tool, API, SDK, CLI or MCP server for a task you can't do yet: describe the task in plain words (job) and get one best_pick with a paste-ready use_it snippet (MCP client config JSON, install command, or API base URL + docs). One call is usually enough: best_pick.summary says in one line what to use and why (reasons: vendor-official or third-party, live check, auth, read-only or can send/spend, outcome reports), best_pick.report_call is the one-line report to send after you use it; results lists ranked alternatives with score breakdowns (evidence, source URL, date), side effects and human steps; verdict is honest (good_match / weak_matches / nothing_good_found / cannot_work_as_asked / not_a_tool_job, then best_pick is null and answer says why; not_a_tool_job adds a curated workaround). Constraints written in the job ('no account', 'nothing paid', 'no new token scopes', 'headless') or in constraints are checked per result (constraint_check). Free, no key. Returns a top-level report_token for optional feedback.",
"inputSchema": {
"type": "object",
"required": [
"job"
],
"properties": {
"job": {
"type": "string",
"maxLength": 500,
"description": "The task in plain words, e.g. 'send SMS to customers', 'read my Google Calendar', 'GitHub MCP server'"
},
"stage": {
"enum": [
"bootstrap",
"connect_auth",
"build_core",
"go_live_get_paid",
"scale_reliability",
"maintain_migrate"
],
"type": "string",
"description": "optional, self-declared"
},
"client": {
"type": "string",
"description": "optional, e.g. claude-code, cursor, my-agent"
},
"detail": {
"enum": [
"full",
"brief"
],
"type": "string",
"description": "brief drops per-result score_breakdown and why[] (about 4x smaller); default full"
},
"agent_type": {
"enum": [
"coding",
"personal_assistant",
"vertical_business",
"commerce",
"research",
"ops_infra",
"orchestrator",
"publisher_directory"
],
"type": "string",
"description": "optional, self-declared"
},
"constraints": {
"type": "object",
"required": [],
"properties": {
"auth": {
"enum": [
"none",
"api_key",
"oauth"
],
"type": "string"
},
"type": {
"enum": [
"mcp_server",
"http_api",
"cli",
"library",
"app",
"directory"
],
"type": "string"
},
"limit": {
"type": "integer",
"maximum": 20,
"minimum": 1
},
"no_paid": {
"type": "boolean"
},
"require": {
"type": "array",
"items": {
"type": "string"
},
"description": "same constraints as a list, e.g. [\"no_account\",\"no_paid\"]; constraints written in the job text are also read"
},
"headless": {
"type": "boolean"
},
"no_admin": {
"type": "boolean",
"description": "no admin install or third-party approval"
},
"read_only": {
"type": "boolean",
"description": "only entries whose every server is verified read-only"
},
"no_account": {
"type": "boolean"
},
"no_new_key": {
"type": "boolean",
"description": "no new API key, token or scopes"
},
"remote_only": {
"type": "boolean"
},
"self_hosted": {
"type": "boolean"
},
"server_side": {
"type": "boolean"
},
"healthy_only": {
"type": "boolean",
"description": "only entries whose latest check passed: MCP handshake ok or auth_required, or (HTTP APIs) base URL reachable"
},
"official_only": {
"type": "boolean",
"description": "publisher domain verified"
}
}
},
"watcher_key": {
"type": "string",
"description": "optional: get an alert if nothing good is found now but a match appears later"
}
}
},
"annotations": {
"readOnlyHint": true,
"openWorldHint": false,
"idempotentHint": true,
"destructiveHint": false
}
}Full Agent Tool Card for one ID (usually not needed: search_tools already returns best_pick.use_it to set it up). Covers: operator, vendor-official vs publisher-verified, capabilities, endpoints and per-client install, auth, side effects (per server for bundles), human steps, blockers, pricing, health history, outcome reports and per-field provenance. Unknown IDs return isError with error_code not_found.
| Parameter | Type | Description |
|---|---|---|
| id * | string | e.g. mcp:com.stripe/mcp; see resolve_tool |
| stage | string | optional, self-declared |
| client | string | optional, e.g. claude-code, cursor, my-agent |
| agent_type | string | optional, self-declared |
{
"name": "get_tool",
"description": "Full Agent Tool Card for one ID (usually not needed: search_tools already returns best_pick.use_it to set it up). Covers: operator, vendor-official vs publisher-verified, capabilities, endpoints and per-client install, auth, side effects (per server for bundles), human steps, blockers, pricing, health history, outcome reports and per-field provenance. Unknown IDs return isError with error_code not_found.",
"inputSchema": {
"type": "object",
"required": [
"id"
],
"properties": {
"id": {
"type": "string",
"description": "e.g. mcp:com.stripe/mcp; see resolve_tool"
},
"stage": {
"enum": [
"bootstrap",
"connect_auth",
"build_core",
"go_live_get_paid",
"scale_reliability",
"maintain_migrate"
],
"type": "string",
"description": "optional, self-declared"
},
"client": {
"type": "string",
"description": "optional, e.g. claude-code, cursor, my-agent"
},
"agent_type": {
"enum": [
"coding",
"personal_assistant",
"vertical_business",
"commerce",
"research",
"ops_infra",
"orchestrator",
"publisher_directory"
],
"type": "string",
"description": "optional, self-declared"
}
}
},
"annotations": {
"readOnlyHint": true,
"openWorldHint": false,
"idempotentHint": true,
"destructiveHint": false
}
}PREFERRED way to give feedback, using the top-level report_token from search_tools/get_tool/compare_tools (covers the first 10 results, valid 30 days). Two kinds: (1) kind=tool_outcome (default): you actually used a tool: tool_id (required when the token covers several tools), outcome worked|partial|failed, used_tool: true (required for tokens from weak/nothing_good_found searches). (2) kind=search_quality: rate the results themselves with verdict results_good|results_partial|results_wrong|missing_tool; never counted against any tool. Note max 300 chars. Reuse your contributor_key.
| Parameter | Type | Description |
|---|---|---|
| kind | string | one of ["tool_outcome","search_quality"] |
| note | string | |
| stage | string | optional, self-declared |
| client | string | optional, e.g. claude-code, cursor, my-agent |
| outcome | string | one of ["worked","partial","failed"] |
| tool_id | string | required for tool_outcome when the token covers more than one tool; never defaulted |
| verdict | string | one of ["results_good","results_partial","results_wrong","missing_tool"] |
| used_tool | boolean | true = you actually called/used this tool |
| agent_type | string | optional, self-declared |
| report_token * | string | |
| expected_tool | string | search_quality: what you expected to find (name or URL) |
| contributor_key | string | Your contributor_key from your FIRST write (format ck_...). Reuse it on every write; omit only on your very first write. We never mint a second key when you send one. |
{
"name": "report",
"description": "PREFERRED way to give feedback, using the top-level report_token from search_tools/get_tool/compare_tools (covers the first 10 results, valid 30 days). Two kinds: (1) kind=tool_outcome (default): you actually used a tool: tool_id (required when the token covers several tools), outcome worked|partial|failed, used_tool: true (required for tokens from weak/nothing_good_found searches). (2) kind=search_quality: rate the results themselves with verdict results_good|results_partial|results_wrong|missing_tool; never counted against any tool. Note max 300 chars. Reuse your contributor_key.",
"inputSchema": {
"type": "object",
"required": [
"report_token"
],
"properties": {
"kind": {
"enum": [
"tool_outcome",
"search_quality"
],
"type": "string"
},
"note": {
"type": "string",
"maxLength": 300
},
"stage": {
"enum": [
"bootstrap",
"connect_auth",
"build_core",
"go_live_get_paid",
"scale_reliability",
"maintain_migrate"
],
"type": "string",
"description": "optional, self-declared"
},
"client": {
"type": "string",
"description": "optional, e.g. claude-code, cursor, my-agent"
},
"outcome": {
"enum": [
"worked",
"partial",
"failed"
],
"type": "string"
},
"tool_id": {
"type": "string",
"description": "required for tool_outcome when the token covers more than one tool; never defaulted"
},
"verdict": {
"enum": [
"results_good",
"results_partial",
"results_wrong",
"missing_tool"
],
"type": "string"
},
"used_tool": {
"type": "boolean",
"description": "true = you actually called/used this tool"
},
"agent_type": {
"enum": [
"coding",
"personal_assistant",
"vertical_business",
"commerce",
"research",
"ops_infra",
"orchestrator",
"publisher_directory"
],
"type": "string",
"description": "optional, self-declared"
},
"report_token": {
"type": "string"
},
"expected_tool": {
"type": "string",
"maxLength": 200,
"description": "search_quality: what you expected to find (name or URL)"
},
"contributor_key": {
"type": "string",
"description": "Your contributor_key from your FIRST write (format ck_...). Reuse it on every write; omit only on your very first write. We never mint a second key when you send one."
}
}
},
"annotations": {
"readOnlyHint": false,
"openWorldHint": false,
"idempotentHint": false,
"destructiveHint": false
}
}Toolwork (slim tool list: search_tools, get_tool, report). When you need a tool, API, SDK, CLI or MCP server for a task you can't do yet, call search_tools({job: '<the task in plain words>'}) first. best_pick.use_it is paste-ready: MCP client config JSON, an install command, or an API base URL + docs. One call is usually enough. get_tool(id) for the full card. Optional: after you use a pick, report({report_token, tool_id, outcome, used_tool: true}) so rankings learn what works. Reads are free with no key; stateless plain JSON (no SSE, no session id required). Every other Toolwork tool (resolve_tool, compare_tools, watch_tools, tell_us, community) stays callable here by name and is listed on the full profile at https://mcp.toolwork.org/mcp/full.
Names, descriptions and schemas were written by the server and are shown as data.