Plan with your assistant

Browse the catalog as a guest

No Adventure Runner profile is needed to list and read the ten named reference routes. Your assistant provider may still require its own account. The guest connection has only list_routes and get_route; it has no personal data, trip-date weather, drafts, editing, approval or exports. Sample schedules and dated seasonal guidance do not confirm current water, permits or access.

1. In your assistant’s custom remote MCP setup, name the connection Adventure Runner Catalog.
2. Use https://mqvrchcbjvnunmuyozlv.supabase.co/functions/v1/mcp/catalog. Choose None or No authentication when prompted; no Adventure Runner OAuth consent is required.
3. Select the catalog in the conversation and send the read-only prompt below. Confirm actual tool use. If your host requires OAuth for every connection, use a host that supports unauthenticated MCP or the public catalog Actions specification.
4. For personal planning features, create or sign in to an Adventure Runner profile and add the account connection at https://mqvrchcbjvnunmuyozlv.supabase.co/functions/v1/mcp. Existing account connections keep working.

• Guest catalog OpenAPI (no authentication)

ChatGPT setup

You connect Adventure Runner to your existing ChatGPT account. Adventure Runner does not ask for your ChatGPT password or an OpenAI API key.

1. In Adventure Runner, sign in and open Account → Connect ChatGPT (or Add a ChatGPT connection). The button copies the MCP server URL and opens ChatGPT Plugins.
2. If developer mode is required, open ChatGPT Settings → Security and login → Developer mode. Availability depends on your account and workspace policy.
3. In ChatGPT Plugins, select the plus button. Name the connection Adventure Runner and paste the copied HTTPS MCP server URL. Use OAuth when prompted for authentication.
4. Complete the Adventure Runner sign-in and consent flow using the account whose plans and preferences you want to use. An existing session may be reused.
5. Start a new ChatGPT conversation and select the Adventure Runner connection from the tools menu. Send the read-only check below. Complete authorization if it is requested on the first tool call.

If ChatGPT does not connect

• No developer mode or plus button: check your ChatGPT account/workspace permissions. Adventure Runner cannot enable a host feature for you.
• The reply has no Adventure Runner tool result: select the connection in this conversation and repeat the read-only check. A conversational answer alone does not verify a connection.
• Authorization fails: check the intended Adventure Runner account is signed in and its assistant access is active. Reopen the connection and complete consent.
• A tool is missing after an update: open the connection in ChatGPT Plugins, choose Refresh, then start a new conversation.
• A trip is missing: share the intended local trip in Adventure Runner Account. A read-only preference check can succeed while your shared-plan list is empty.

1. Choose a route

Start with a comparison. Inspect the reference variant, geometry, sources, season, water candidates and logistics before selecting it. Comparing public references does not share a personal trip.

2. Check conditions for your date

Choose the exact route variant returned by the assistant. Request trip-date weather with a readable refresh time and timezone. Seasonal or historical context is not a forecast. Fire, smoke, closures and trail access remain unchecked by the service; use official notices for those checks.

3. Build a draft from your answers

The assistant needs the route variant, date, local start time, total moving hours and day-trip versus reference-camps style. It should ask before proposing a draft if any required answer is missing. Reference camps are unreserved and their sample itinerary resumes at 8:00 AM. A successful draft response includes a proposal identifier and pending owner approval; it has not saved a plan.

4. Review and approve in Adventure Runner

Return to Adventure Runner → Account → Refresh access and proposals. Match the proposal identifier, then inspect its route, date, timing, weather, packing and unresolved arrangements. Approve only after reviewing the details, or reject it. Approval creates a shared account snapshot. Host tool permission does not perform this owner approval.

To edit in the browser planner, choose Save as local copy. Later local changes do not automatically update the shared snapshot; share the intended local trip again when you want the assistant to use those changes.

5. Request a PDF and GPX

After approval, ask the assistant to identify the approved shared plan and export that plan. The PDF includes the itinerary, gear and preparation checklist, source context and weather refresh time when available. GPX requires mapped geometry. A pending draft can also be exported, but its PDF remains marked pending approval.

Download links are valid for up to 24 hours and anyone who has a link can download its file. Account → Revoke download links stops future downloads. Saved revision changes, stopping sharing or reviewing a draft also invalidate its links. Request a new export if needed; downloaded copies cannot be recalled.

6. Finish preparation

Open the local copy in the planner to check Packing and Before you go. Confirm water refills, permits/access, campsites where relevant, transport and your emergency/check-in plan. The service does not book permits or reservations. Download maps and test offline reopening on the device you will carry. Use Share → Download trip for an editable JSON backup; PDF and GPX are different exports.

Personalize your gear

Account gear rules are shared across authorized assistants. “Below” excludes the threshold itself; gloves below 55°F applies at 54°F, not 55°F.

Set weather review triggers

These are example thresholds to replace with your own values. Forecast assessments run on request; there is no background monitoring or notification service. Missing, stale, seasonal or historical data leaves the assessment unknown.

Synaptic Chat: connect through Actions

Tested integration: OAuth authorization, route catalog read, account preference read and a pending preference proposal were verified with owner review and rejection. Draft creation and PDF/GPX export are exposed by the same Actions API but have not been tested end to end in Synaptic Chat.

1. Open Synaptic Chat and select an Adventure Runner agent configured and made available by your host operator. If none is available, ask the operator for access; pasting the MCP URL into a normal conversation does not install an agent.
2. For the first connection, send only the read-only preference prompt below. Finish its sign-in prompt before sending other tool requests.
3. Authorize the Adventure Runner account whose plans and preferences you want to use. An existing Adventure Runner session may be reused.
4. Confirm a tool result returns your saved values or explicitly identified defaults. Then continue with route questions or the trip walkthrough.

If Synaptic Chat stalls

• No Adventure Runner agent: the host operator must share or install the configured agent for your account. Users cannot complete that step by pasting credentials or a schema into chat.
• Several first requests wait for sign-in: complete one preference-read authorization, then retry requests one at a time.
• The result belongs to a different account: stop and check the Adventure Runner identity used for consent before reading or sharing personal plans.
• A proposed change has not been saved: open Adventure Runner Account, refresh proposals and approve or reject the matching identifier. Chat approval alone does not save it.

Synaptic Code: use enabled tools

Limited verification: Adventure Runner tools were available in one tested harness session and a public-route read succeeded. Account authorization, draft approval and file export have not been verified end to end in Synaptic Code. Availability is not automatic for every user.

1. Check whether your Synaptic Code session exposes Adventure Runner tools. If not, ask the host operator to enable a compatible remote MCP connection.
2. Ask the assistant to list the Adventure Runner reference routes using its tools. Confirm the reply is a tool result rather than a general answer.
3. For personal preferences or shared plans, complete Adventure Runner authorization if requested and try the same read-only preference check used above. Do not assume a successful public-route read proves personal account access.
4. Once your account read succeeds, follow the trip walkthrough. Review and approve proposals in Adventure Runner; report any unsupported host step rather than treating it as completed.

If Code tools are unavailable

Ask the host operator to check the session’s MCP connection and permissions. A working connection in Synaptic Chat does not automatically enable the Code harness. If personal account access fails, stop at the read-only check and confirm the host supports the required OAuth flow.

• Actions exposes the same eleven operations and approval rules as MCP: routes, weather, shared plans, preferences, proposals and PDF/GPX exports. Each authorized call uses that user’s Adventure Runner account.
• Host client credentials belong in host/server secret stores. Users authorize their own accounts; they do not share an agent’s client secret.

Anthropic: connect Claude

Use Claude’s custom remote connector on the web or desktop. Team or Enterprise administrators may need to add the connector before members can connect.

1. Open Customize → Connectors → + Add → Add custom connector. Enter Adventure Runner and the MCP URL above.
2. Choose authenticated access. If asked how to identify the OAuth client, choose Register automatically; Adventure Runner supports dynamic client registration.
3. Finish adding the connector, select Connect or Sign in now, and complete Adventure Runner sign-in and consent for your account.
4. In a conversation, open + → Connectors and enable Adventure Runner. Run the read-only check below before starting the trip walkthrough.

• Claude custom connector instructions

If Claude cannot connect

Missing Add custom connector: check your account or ask your workspace administrator. A connected account without tool results usually needs the connector enabled in this conversation. If authentication fails, confirm Register automatically and repeat consent; do not substitute a database key.

xAI: connect Grok

Personal Grok accounts can add custom connectors directly. Business and Enterprise additionally require an administrator to provision the connector. Google sign-in, account connection and read-only preferences/routes were verified with Adventure Runner on October 5, 2026; voice and the full draft/export workflow remain unverified.

1. For a personal account, open grok.com/connectors → New Connector → Custom. Name it Adventure Runner and enter the account MCP URL above (or the guest catalog URL for read-only route exploration).
2. Complete Adventure Runner OAuth for the account connection; the guest catalog needs no Adventure Runner authentication.
3. For Business/Enterprise, a team administrator first uses console.x.ai → Grok Business → Connectors → Add Connector → Other. Team members then connect their own accounts at grok.com/connectors.
4. Select the connector for your conversation and request the read-only check below.

• Grok team connector management

Grok Build CLI alternative

With Grok Build installed and signed in, add the remote HTTP server using the commands below. Complete browser OAuth when prompted; the /mcps panel also offers authentication with i. This is a terminal workflow.

grok mcp add --transport http adventure-runner https://mqvrchcbjvnunmuyozlv.supabase.co/functions/v1/mcp
grok mcp list
grok mcp doctor adventure-runner

• Grok Build MCP setup

If Grok cannot connect

For personal accounts, use New Connector → Custom at grok.com/connectors. Missing Business/Enterprise team controls: ask an administrator with team management permission, or use the CLI path. If tools are absent, verify that you connected your personal account after team provisioning. In Grok Build, run the doctor command and complete OAuth before repeating the read-only check.

Google: connect Gemini

If Custom apps is available, set up the connection in Gemini on the web. Google currently limits this feature to personal accounts, adults in the US, English, and Keep Activity enabled; work or school accounts are excluded.

1. Open gemini.google.com → Settings → Connected Apps. If needed, open Personal Intelligence first.
2. Under Custom apps, choose Add a custom app and paste the MCP URL above. Adventure Runner supports dynamic client registration; leave manual advanced client credentials unset.
3. Choose Next and complete the account authorization prompts using your Adventure Runner account.
4. In chat, type @ and select the connected app. Run the read-only check below. Google also documents use on mobile after web setup.

• Google custom app setup and eligibility

Gemini CLI alternative

With Gemini CLI installed and signed in, run the terminal command below. Then enter /mcp auth adventure-runner inside Gemini CLI and complete browser consent. Use /mcp to inspect available tools, then send the same read-only preference check.

gemini mcp add --scope user --transport http adventure-runner https://mqvrchcbjvnunmuyozlv.supabase.co/functions/v1/mcp

• Gemini CLI MCP and OAuth instructions

If Gemini cannot connect

Missing Custom apps: check Google’s eligibility requirements and account settings, or consider Gemini CLI. A normal Gemini conversation without a selected connected app does not verify tools. If OAuth fails, record the failed step and host version for troubleshooting; do not paste Supabase keys into the chat or CLI configuration.

Ollama: use a local model through OpenCode

Ollama runs the model; a compatible client handles remote MCP and account authorization. This setup uses OpenCode. Local inference still requires online Adventure Runner tool calls and weather access.

1. Install Ollama and OpenCode using their official instructions. Run ollama launch opencode and select a local model that supports tool use with enough context for the tool definitions.
2. Merge the mcp entry below into your existing OpenCode configuration, commonly ~/.config/opencode/opencode.json. Keep your existing provider and model settings.
3. Run opencode mcp auth adventure-runner and authorize your Adventure Runner account in the browser. Use opencode mcp list to check the connection.
4. Start OpenCode with your Ollama model and request the read-only check below. Confirm an Adventure Runner tool result before building or changing a plan.

{
  "$schema": "https://opencode.ai/config.json",
  "mcp": {
    "adventure-runner": {
      "type": "remote",
      "url": "https://mqvrchcbjvnunmuyozlv.supabase.co/functions/v1/mcp",
      "enabled": true
    }
  }
}

opencode mcp auth adventure-runner
opencode mcp list

• Ollama’s OpenCode setup
• OpenCode remote MCP configuration and OAuth

If Ollama or OpenCode cannot connect

A model answer without a tool call is not a connection check. Confirm OpenCode lists the server and that the selected model supports tool use. Run opencode mcp debug adventure-runner to inspect connection and OAuth failures. Plain Ollama chat does not automatically provide remote MCP authorization.

If another assistant cannot connect

A host without remote MCP/OAuth or compatible Actions cannot use these account tools through a plain chat. You can still plan in Adventure Runner and share a PDF with your assistant for discussion. That file-based workflow does not give the assistant live account access or a way to save preference or plan changes.