thegoodweb.ai
Launching Soon
https://thegoodweb.ai/Connect straight to this server’s public endpoint.
https://mcp.thegoodweb.ai/mcpWe add this server to your workspace, then open Studio — saved access, one connection to many servers, with a history of what ran.
Last probed Sep 14, 2026 · mcp.thegoodweb.ai
26tools discovered
Showing 25 of 26 from the live probe.
Open The Good Web
Opens The Good Web in a panel: the live directory of creator agents, example questions, and an ask box the user types into directly. FREE - nothing is asked and no Web Credits are spent. Call this when the user says "open The Good Web", "show me The Good Web", or asks what the network is, and nothing else. It does not answer questions; ask_the_good_web is what answers.
List The Good Web agents
Returns the directory of verified agents on The Good Web. Call this BEFORE ask_agent when the user names a person, publication, or community but you do not know their handle, or when the user asks for a recommendation among agents. Each result includes the agent's handle and display name; users can refer to an agent by either form. Filter by `flank` to narrow the cast (person = individual creators, publication = editorial brands, community = live feed graphs). Set `verifiedOnly: false` only when
Ask a specific Good Web agent
Asks a SINGLE named agent on The Good Web a question and returns that agent's reply with citations. Costs 10 Web Credits (one agent). PREFER this tool over ask_the_good_web whenever the user's prompt names a specific creator, publication, or community. Common patterns that should route here, NOT to ask_the_good_web: 'ask Sequoia about X', 'What does Sequoia think of X', 'Have Claus answer this', '@sequoia X', 'Section Hiker, what about Glacier Peak'. The presence of a single agent name in the qu
Choose which Good Web agents should answer
Runs only the classifier that picks which verified agents on The Good Web would answer a question, and returns them without asking any of them. FREE: no agent runs, no Web Credits are spent, and no confirmation is needed. This is a preview, not an answer - it exists so a caller can show who is about to be asked while the fan-out runs. Pass the returned handles straight to ask_the_good_web so the classifier decides once and the names shown are the ones that answer. Do NOT call this on its own exp
One creator's answer in a Good Web fan-out
ONE STEP of an incremental network fan-out, for a caller that renders each creator's answer as it arrives. Asks a single named creator and charges 10 Web Credits for that one answer. DO NOT CALL THIS TO ANSWER A USER'S QUESTION. It answers as one creator with no cross-creator comparison, so used on its own it gives a partial view of the network. Use ask_the_good_web for the network's answer, or ask_agent when the user named one specific creator; both cost the same per creator and return a compl
Compare the answers a Good Web fan-out returned
Writes the SHORT comparative header over answers a fan-out has already returned: where the creators converge, named and counted, and where they genuinely disagree. FREE - the per-creator steps already charged for the asking. Send the answers you received from good_web_fanout_step. This does not ask anyone anything and does not restate the answers; it says only what is invisible from any single one of them. Call it after the per-creator steps have returned, never instead of them.
Open a creator's Good Web profile
Returns one creator's Universal Profile on The Good Web: their display name, what their corpus covers, how many pieces of their work are indexed, their most recent titles, and the 10 Web Credits it costs to ask them. FREE - this reads their record and asks them nothing. Use it when the user wants to know about a creator rather than to put a question to them; ask_agent is what asks.
Ask The Good Web (network-wide aggregation)
Fans the question out across the verified agents on The Good Web. A fast classifier picks the 3-5 verified agents most likely to have substantive coverage based on their topics blurbs, then those agents answer in parallel. Returns per-agent attribution plus a synthesis and a combined sources rail (web articles + podcast clips). Spends 10 Web Credits per agent that returns a successful reply, capped at 40 credits per query. DO NOT call this tool when the user named one specific agent in their pro
Ask the Curator
Ask the Curator a question about the reader's edition or the news it covers, and get a sharp editor's answer with linked source cards. This is the front door for "go deeper" on anything in today's edition: the Curator cites the reader's own edition articles alongside fresh web results. PREFER this tool for edition questions, background, and "what's the story with X" - it does NOT fan out to other agents. To have a specific claim checked against sources, use fact_check instead. Scope the questio
Check a claim
Look up a claim against live web sources and report what they say - supported, mostly supported, needs context, mostly contradicted, contradicted, or could not verify - with a short explanation and linked sources. Use this when the reader wants a claim checked, not for open-ended background (use ask_the_curator for that). This reports what sources say rather than issuing a ruling, so present it that way. Costs 8 Web Credits. To check a STORY IN THE READER'S EDITION, pass its id as `item` and om
Summarize a story in the edition
Read one story in the reader's Good Web Daily and return a short overview of it: a two-or-three sentence TLDR plus 3-5 key points, taken from the article's own text with no web search. Costs 2 Web Credits, and NOTHING when the article cannot be read (a paywall or a consent wall) - offer ask_the_curator in that case. For background, context, or anything the article does not itself answer, use ask_the_curator instead. When the reader's message names an id - for example `item "it-c12"` or `section
Suggest follow-up questions for a story
Internal helper invoked by the Curator widget itself when the reader opens a story's ask panel: returns 2-3 suggested follow-up questions about that story. Free. The assistant must NEVER call this - it fills in buttons inside the widget and answers nothing. To actually answer a question about a story, call ask_the_curator with that story's `item` id.
Suggest follow-up questions for an answer
Internal helper the panel invokes about an answer it has already rendered: returns up to 3 suggested follow-up questions for the reader to continue with. Free, asks nobody anything and changes nothing. The assistant must NEVER call this - it fills in buttons inside the panel and answers nothing. To actually answer a follow-up, call ask_agent with that creator's handle.
Show the user's Good Web Daily
Returns the user's daily edition and renders it as an interactive card: an immersive cover, local weather, and curator-hosted sections (top stories, stocks, deep dives, local, sports) with real images and links. Call this when the user asks to SEE, get, read, open or show their edition - "my edition", "my daily edition", "today's edition", "the daily briefing", "my morning edition", "the news", or "what's happening today". Asking to SEE the edition belongs here and is free; only a question ABOUT
Refresh the daily edition
FREE same-day refresh of the user's Good Web Daily: stories they have already read are swapped for fresh unseen ones (marked New); the rest of the edition stays stable. Call when the user asks to refresh or update their edition. For a full re-composition with new sections and themes, use recompile_edition instead (that one costs credits).
Resume a paused edition
FREE. Composes today's edition for a reader whose edition is PAUSED - call this only after get_edition or refresh_edition has reported it paused. Takes one to two minutes. Costs no credits, so it needs no spending confirmation, but do tell the user it is starting because they will wait for it. Not a substitute for recompile_edition, which re-composes an edition that is already current and costs credits.
Re-compose the daily edition (paid)
PAID full re-composition of the user's Good Web Daily: a fresh editorial pass with new section structure and themes, drawn from the current pool with never-reshow applied. Costs 25 credits and takes one to two minutes. Two-call shape: call once without confirmationToken to get the exact cost and a confirmation token, relay the cost to the user and get their explicit go-ahead, then call again with confirmationToken to actually spend the credits and run it. For a free same-day update, use refresh_
Open a premium story from the edition
Unlocks and returns the full text of a premium story in the user's daily edition, charging the small per-read credit cost to the user and crediting the publisher. Idempotent: re-opening the same story does not charge again. The edition card calls this when the user opens a premium story; you normally do not call it directly.
Update daily edition preferences
Change what appears in this reader's future editions. Call this whenever they ask for a change - it IS the change; nothing else records anything, so never say you have noted something without calling it. Free, and it never draws credits. Changes shape the NEXT edition, not the one they are reading. Send every distinct change as its own action in one call. Do NOT call this for things it cannot do: "put technology second and sports third" (pin one section to lead, and turn the others up so they pl
Record edition card feedback
Internal telemetry beacon invoked by the Curator widget's own buttons (more-like / less-like / share). Records the signal for edition personalization and returns a one-line confirmation. The assistant should NOT call this on the user's behalf - when the user states a preference in conversation, use update_edition_preferences instead, which records durable preferences with rank weights.
Record edition reading telemetry
Internal telemetry beacon invoked by the Curator widget itself as the reader scrolls, dwells on cards, opens links, and hands a contextual question to the conversation. Batched and fire-and-forget. The assistant must NEVER call this - it reports what the widget observed, not anything the user asked for. To act on a stated preference use update_edition_preferences; to record an explicit tune use record_edition_feedback.
Record agent-card click telemetry
Internal telemetry beacon a widget invokes about itself. Reports twelve kinds of observation: a citation click, another outbound link click, a Content-Security-Policy refusal inside the widget's own sandbox, a request for more room, a contribution-disclosure toggle, a qualified audio play the widget's own player measured, a click on a section's topic link, a conversation-strip toggle, a click on a strip's Surf link, a creator dropped from or restored to a fan-out before it is paid for, a press o
Save an edition story for later
Saves a story, video, or podcast episode from the user's daily edition to their saved-for-later list. Call when the user asks to save, bookmark, or keep an edition item (the widget's bookmark button also calls this). The item is resolved server-side from the edition, so only editionId and itemId are needed. Saving is idempotent per story.
Remove an edition story from saved
Removes a story, video, or podcast episode from the user's saved-for-later list. Call when the user asks to unsave, un-bookmark, or remove an item they previously saved from their daily edition (the widget's bookmark button also calls this to toggle a saved card off). The item is resolved server-side from the edition, so only editionId and itemId are needed.
List saved edition stories
Returns the user's saved-for-later list from their daily editions (newest first): stories, videos, and podcast episodes they bookmarked. Call when the user asks what they saved, wants to read something from their list, or asks for their reading list.
Get your MCP into directories
A working endpoint is step one. Directory coverage is the coordinated launch across ChatGPT, Claude, Cursor, the MCP Registry, and community indexes.
Directory coverage for brandsLaunching Soon
Use the MCP endpoint listed on this page in your MCP client configuration. One-click install pills support Claude, Cursor, VS Code, and other hosts. Copy the remote MCP URL if your client needs a manual entry.
Operate Thegoodweb? Verify ownership to take over this directory entry.
This server appears in the MCPBundles directory. Verify you operate it to take over the listing — name, description, logo, contact email, and skill content. We email a 6-digit code to a maintainer address your server publishes in /.well-known/security.txt or /.well-known/mcpbundles.json. Free, takes about a minute.
MCPBundles probed 26 tools on the live server. The tool list on this page reflects what was discovered at the last refresh — connect your client to see the full set available to your session.
No provider sign-in was required during MCPBundles' probe. Your client may still need MCPBundles credentials depending on how you connect.
MCPBundles is an independent platform built on the open Model Context Protocol standard. Not affiliated with Anthropic PBC or Claude.