MCP server

https://mcp.gridrouter.io/mcp serves the same operations as the REST API. Each tool is one operation: the same input schema, the same result, the same errors. Tools a connection's scopes do not cover are not listed; public tools such as catalog_search are always listed.

Every tool result includes the call's cost in _meta["io.relaygrid/cost_micro"]. A failed call returns isError: true with code: message text and the error body as structured content.

ResourcesCopy Anchor Link

Tools that start a job or list return a resource_link to its grid:// URI.

PromptsCopy Anchor Link

• find_verified_email (first_name, last_name, domain): runs people.email.find, then email.verify on the result, and reports the total cost.
• research_account (domain, budget_usd, default 0.50): quotes each call and keeps the research under budget.

ToolsCopy Anchor Link

The same list is available as JSON from GET https://api.gridrouter.io/v1/tools?format=mcp (or openai / anthropic for function-calling schemas).

Tool	REST	Scope	Description
catalog_search	GET /v1/catalog/endpoints	public	Find vendor endpoints by keyword (capability, vendor, what they return). Returns price and quality.
catalog_get	GET /v1/catalog/endpoints/{provider}/{endpoint}	public	Full definition of one endpoint: input JSON Schema, pricing model and quality.
vendor_search	GET /v1/catalog/vendors	public	Every GTM data vendor GridRouter knows about (integrated or researched): API access, spec status, published rate limits and marketplace listings.
vendor_get	GET /v1/catalog/vendors/{slug}	public	One vendor's sourced record: company, API surface, docs, pricing, data, compliance, relationships and marketplace listings, each with sources and confidence.
vendor_openapi	GET /v1/catalog/vendors/{slug}/openapi	public	The stored OpenAPI document for a vendor: the vendor's official spec, or one GridRouter reconstructed from its docs (x-reconstructed, x-confidence, x-source-url on every operation).
vendor_rate_limits	GET /v1/catalog/vendors/{slug}/rate-limits	public	Published rate limits (per window, concurrency, headers, 429 behaviour) with source and check date, and the pacing GridRouter derives from them.
capabilities_list	GET /v1/catalog/capabilities	public	Capabilities (normalized tasks like people.email.find) and the endpoints that serve each.
categories_list	GET /v1/catalog/categories	public	Top-level GTM data categories.
leaderboard_get	GET /v1/catalog/leaderboards/{capability}	public	Endpoints for a capability ranked by usage-derived quality (Wilson lower bound).
quote	POST /v1/quote/{provider}/{endpoint}	call	Worst-case cost of calling an endpoint with this input, without calling it.
call	POST /v1/call/{provider}/{endpoint}	call	Call one vendor endpoint. Reserves the worst-case cost, charges the actual cost. Async vendors return a job. Runs as an MCP task.
run_auto	POST /v1/run/auto	run	Auto router: GridRouter picks the capability from the input fields you have (and the output fields you ask for), then routes it like run. The chosen capability and the alternatives are returned.
run	POST /v1/run/{capability}	run	Routed call: GridRouter picks vendors (waterfall/cheapest/best) until one hits. Only hits are charged on per-success vendors.
call_get	GET /v1/calls/{id}	call	Cost, timings (total, upstream, overhead, TTFB, queue), provider, outcome and every routed attempt of one call, like OpenRouter's generation lookup. Keys without logs:read only see calls they made. Available a few seconds after the call.
presets_list	GET /v1/presets	catalog:read	Saved routing presets for this org: provider preferences, field selection, max cost and stop rule. Use one as preset: "@slug" on run.
presets_get	GET /v1/presets/{slug}	catalog:read	One saved routing preset.
presets_create	POST /v1/presets	presets:write	Create or replace a routing preset shared within the org (order/only/ignore/sort/max price/data policy, credential, fields, max cost, stop rule).
presets_delete	DELETE /v1/presets/{slug}	presets:write	Delete a preset. Requests that name it fail with not_found afterwards.
job_create	POST /v1/jobs	jobs	Run any endpoint in the background through the job queue, with per-vendor pacing. Runs as an MCP task.
job_get	GET /v1/jobs/{id}	jobs	Status and result of a queued or async job.
list_create	POST /v1/lists	lists	Run an endpoint or capability over up to 10,000 rows in the background. Runs as an MCP task.
list_get	GET /v1/lists/{id}	lists	Progress and cost of a list run.
list_rows	GET /v1/lists/{id}/rows	lists	One page of per-row results for a list.
balance_get	GET /v1/balance	balance:read	Prepaid balance, amount held for in-flight calls, and what is available.
budget_set	PUT /v1/budgets	keys:write	Cap spend for the org, one key, or one agent per day, month or total.
keys_list	GET /v1/keys	keys:write	API keys for this org (never the secret).
keys_create	POST /v1/keys	keys:write	Mint a scoped key. The secret is returned once.
keys_revoke	DELETE /v1/keys/{id}	keys:write	Revoke a key. Billable calls stop immediately.
recovery_codes_create	POST /v1/recovery-codes	keys:write	Owner only (a * key): issue 10 one-time recovery codes that can mint a new owner key if every key is lost. Replaces earlier codes; shown once.
keys_rotate	POST /v1/keys/{id}/rotate	keys:write	Mint a replacement key with the same scopes and limits; the old key keeps working for grace_seconds (default 24 h), then expires.
keys_limit_set	PUT /v1/keys/{id}/limit	keys:write	Cap what one key can spend, resetting daily, weekly, monthly or never. null removes the cap. Calls over the cap fail with budget_blocked.
agents_create	POST /v1/agents	agents:write	Create an agent with its own budget and mint an agent key for it.
log_views_list	GET /v1/logs/views	logs:read	Your saved log views plus views shared with the workspace, newest first. A view is a query in the log query language and the visible columns.
log_views_create	POST /v1/logs/views	logs:read	Save the query and visible columns under a name. shared: true (owners and admins only) makes it visible to the whole workspace; replace: true overwrites your view with the same name.
log_views_update	PUT /v1/logs/views/{id}	logs:read	Rename a view or change its query, columns or sharing. You can change your own views; owners and admins can also change shared ones.
log_views_delete	DELETE /v1/logs/views/{id}	logs:read	Delete one of your views (owners and admins can also delete shared views).
log_views_import	POST /v1/logs/views/import	logs:read	One-time move of views and pins the dashboard kept in the browser. Views whose names you already use and calls already pinned are skipped.
log_pins_list	GET /v1/logs/pins	logs:read	Calls you pinned to the top of the log viewer, newest pin first, each with its current log row (null once the call is past log retention).
log_pins_create	POST /v1/logs/pins	logs:read	Pin a call to the top of the log viewer. You keep up to 20 pins; the oldest drops off.
log_pins_delete	DELETE /v1/logs/pins/{call_id}	logs:read	Remove a pinned call.
logs_list	GET /v1/logs	logs:read	Every call and routed attempt for this org, newest first (360-day retention). Filter by time, outcome, status, provider, capability, endpoint, key, agent, client, mode, credential, error code, latency and cost; cursor-paginated.
logs_facets	GET /v1/logs/facets	logs:read	Counts per value (outcome, status, provider, capability, endpoint, key, client…) and latency/cost ranges for the same filters as logs_list.
logs_stats	GET /v1/logs/stats	logs:read	Calls, error rate, p50/p95/p99 latency, spend, a hit/miss/failed timeline and spend by provider over time, for the same filters as logs_list.
logs_billing	GET /v1/logs/billing	logs:read	Per vendor, how many of your own-key (BYOK) calls were billed as the vendor's published rule says, how many were charged for a miss, double-charged on a retry or at the wrong price, and the credits and dollars at stake. Same filters as logs_list.
logs_get	GET /v1/logs/{id}	logs:read	One call with timings, its redacted request/response body (audit-logged), the other attempts in its routed run or list, and its ledger entries.
settings_get	GET /v1/settings	logs:read	Org data settings, such as whether request/response bodies are stored.
settings_update	PUT /v1/settings	keys:write	Turn body storage on or off (off stops new bodies from being kept) and set how many days of call log members and keys can read (30, 90, 180 or 360).
credentials_list	GET /v1/credentials	credentials:write	Your own vendor keys (BYOK) stored for this org: provider, slot, label, field names, the key's last four characters and its last test result. Secrets are write-only and never returned.
credentials_put	PUT /v1/credentials/{provider}	credentials:write	Store your own key for a provider as the primary or fallback slot, envelope-encrypted. BYOK calls are never charged by GridRouter.
credentials_remove	DELETE /v1/credentials/{provider}	credentials:write	Delete one slot of your own key for a provider. Calls fall back to the next slot.
credentials_test	POST /v1/credentials/{provider}/test	credentials:write	Call the provider's free, read-only test endpoint (for example a credit balance) with your stored key. Never charged; the call is logged.
credentials_verify	POST /v1/credentials/{provider}/verify	credentials:write	Call the provider's free, read-only test endpoint with a key that is not stored yet. Nothing is saved; the key is used for this one call. Never charged; the call is logged.
credentials_update	PUT /v1/credentials/{provider}/label	credentials:write	Set or clear the label of a stored key. The secret is not touched.
credentials_swap	POST /v1/credentials/{provider}/swap	credentials:write	Make the fallback key the primary and the primary the fallback for one provider. Takes effect on the next call.
apps_list	GET /v1/apps	logs:read	Apps attributed on your calls (X-Grid-App, or HTTP-Referer + X-Title like OpenRouter) with 30-day usage, plus how each is listed.
apps_put	PUT /v1/apps/{id}	keys:write	Name an attributed app and opt it in (or out) of the public rankings and showcase. A domain-style id must match its URL; another org's public id is refused.
auth_code_create	POST /v1/auth/codes	keys:write	Consent step of Sign in with GridRouter (OAuth PKCE, S256): a signed-in member approves an app and gets a one-time code (10 minutes) to redirect back with.
auth_keys_exchange	POST /v1/auth/keys	public	Final step of Sign in with GridRouter: exchange the one-time code and its PKCE code_verifier for a user-controlled key with call/run scopes and any credit limit the user set.
rankings_get	GET /v1/rankings	public	Top capabilities, providers, categories and opted-in apps by calls and hits over the last day, week or month across all GridRouter traffic, with change vs the previous window and trending movers.
apps_directory	GET /v1/catalog/apps	public	Public apps built on GridRouter (owners opted in), ranked by calls with their top capabilities and providers.
app_profile_get	GET /v1/catalog/apps/{id}	public	One public app: 30-day usage by day, top capabilities and providers.
activity_get	GET /v1/catalog/activity	public	Per-endpoint uptime (share of calls without a vendor-side failure), p50/p95 latency and volume over time for a capability or provider, plus the top public apps using it.
status_get	GET /v1/status	public	Live health per provider: vendor failures in the last 30 seconds (the router's outage signal), last-hour error rate and latency, and 24-hour and 30-day uptime with daily bars.
tools_export	GET /v1/tools	public	Endpoint and capability tools in OpenAI, Anthropic or MCP format.
registry_validate	POST /v1/registry/validate	public	Lint a provider + endpoints YAML submission against the registry schema and margin guard.
usage_by_tag	GET /v1/usage	balance:read	Spend and calls grouped by a meta tag over a window (reads the Postgres mirror). (planned: not yet available)
registry_submit	POST /v1/registry/submissions	keys:write	Submit catalog YAML (validated as in registry_validate) for review by the GridRouter team.
announcements_list	GET /v1/announcements	public	Current platform announcements (maintenance windows, incidents, launches).
review_create	POST /v1/reviews/{provider}/{endpoint}	call	Leave a rating backed by your verified call count. (planned: not yet available)
waterfall_catalog	GET /v1/waterfalls/catalog	pipelines:read	Every capability with its inputs, normalized output fields, and the endpoints a waterfall can use: price, quality, fill, p50 and rate limits.
waterfall_templates	GET /v1/waterfalls/templates	pipelines:read	Built-in starting points (work email, mobile finder, company enrich + technographics, and sandbox versions).
waterfall_list	GET /v1/waterfalls	pipelines:read	Waterfalls and pipelines in this org, newest first, with their published version.
waterfall_create	POST /v1/waterfalls	pipelines:write	Create a draft waterfall (one capability, ordered vendor steps, speed profile, stop rules, per-field merge) or a multi-stage pipeline. Start from a definition, a template or a routing preset.
waterfall_get	GET /v1/waterfalls/{id}	pipelines:read	One waterfall: its draft definition, versions and published endpoint.
waterfall_update	PUT /v1/waterfalls/{id}	pipelines:write	Replace the draft definition, title or description. Published versions never change.
waterfall_delete	DELETE /v1/waterfalls/{id}	pipelines:write	Delete a waterfall and its versions; its published endpoint stops answering.
waterfall_estimate	POST /v1/waterfalls/estimate	pipelines:read	Per-step quote, p50 latency, fill and reach, plus worst-case and expected cost and ETA for a definition, without calling any vendor.
waterfall_versions	GET /v1/waterfalls/{id}/versions	pipelines:read	Every published version (immutable), newest first.
waterfall_publish	POST /v1/waterfalls/{id}/versions	pipelines:write	Freeze the draft as a new semver version and serve it at /v1/x/{workspace}/{name} (REST, MCP x_run and OpenAPI).
waterfall_rollback	POST /v1/waterfalls/{id}/rollback	pipelines:write	Serve an earlier published version again. Nothing is deleted.
waterfall_preset	GET /v1/waterfalls/{id}/preset	pipelines:read	The waterfall's vendor order as a routing preset body for presets_create, so /v1/run can use it as @slug.
waterfall_openapi	GET /v1/waterfalls/{id}/openapi	pipelines:read	OpenAPI 3.1 for the waterfall's published endpoint /v1/x/{workspace}/{name}.
waterfall_run	POST /v1/waterfalls/{id}/run	pipelines:run	Run a waterfall on one input. Sync by default; Prefer: wait=N (or wait) falls back to 202, Prefer: respond-async (or async: true) returns 202 with a run to poll, stream (/events) or receive by webhook. version draft runs the test runner.
waterfall_batch_run	POST /v1/waterfalls/batch-run	pipelines:run	Run a waterfall over up to 1,000 inputs (JSON rows or CSV) in the background through gr-rows, paced per vendor. Returns the batch (a list) to poll. Runs as an MCP task.
waterfall_runs	GET /v1/waterfalls/{id}/runs	pipelines:read	Recent runs of one waterfall: outcome, cost, time, attempts and fields filled.
waterfall_status	GET /v1/waterfalls/{id}/runs/{run_id}	pipelines:read	One run: status, golden record with _provenance per field, every attempt with its timing and error, cost per filled field.
waterfall_run_events	GET /v1/waterfalls/{id}/runs/{run_id}/events.json	pipelines:read	Progress events of a run after a sequence number (the SSE stream's JSON form).
waterfall_cancel	POST /v1/waterfalls/{id}/runs/{run_id}/cancel	pipelines:run	Stop a queued or running run. In-flight attempts are aborted and released; the partial result is kept.
x_run	POST /v1/x/{workspace}/{name}	pipelines:run	Call your own published waterfall by name: /v1/x/{workspace}/{name}, where workspace is your org id. Same modes as waterfall_run.
plans_list	GET /v1/plans	public	Every GridRouter plan with its price, limits, features and managed-call markup, plus the shared overage rates and always-free list.
entitlements_get	GET /v1/entitlements	balance:read	This org's plan, resolved limits (null means no cap), features and usage this month, including logged calls and overage blocks.
billing_get	GET /v1/billing	balance:read	Billing mode, plan and subscription status, balance (with any simulated test credit), auto-recharge and receipts.
billing_checkout_create	POST /v1/billing/checkout	billing:write	Buy credits (amount_usd, the 5.5% top-up fee is added) or upgrade to a paid plan. Returns a URL: Stripe Checkout, or the test purchase page when billing is simulated.
billing_checkout_get	GET /v1/billing/checkout/{id}	balance:read	One checkout session and its status (open, completed, declined, requires_action).
billing_checkout_complete	POST /v1/billing/checkout/{id}/complete	billing:write	Simulated billing only: approve, decline or require card action on a test checkout. Emits the same events Stripe would; no card is charged.
billing_plan_change	PUT /v1/billing/plan	billing:write	Move to a lower plan, or to Free to cancel. Upgrades go through billing_checkout_create.
billing_autorecharge_set	PUT /v1/billing/auto-recharge	billing:write	Top up by amount_usd whenever the balance drops below threshold_usd.
billing_simulate	POST /v1/billing/simulate	billing:write	Test and simulated modes only: payment failure, dispute (freezes the org), refund, renewal, downgrade, cancel, auto-recharge, grant expiry, or reset simulated state.
options_schema	GET /v1/options/schema	public	JSON Schema of ExecutionOptions: timing, concurrency, rate limits, cost, routing, data and cache, delivery, identity and observability. Every request body accepts it as options.
options_resolve	POST /v1/options/resolve	catalog:read	What a request would run with: request options over the preset, your key's defaults and the workspace defaults, clamped by your plan. Returns where each value came from and what was clamped.
options_defaults_get	GET /v1/options/defaults	logs:read	The workspace's default ExecutionOptions (the lowest precedence layer).
options_defaults_put	PUT /v1/options/defaults	presets:write	Replace the workspace's default ExecutionOptions (cache TTLs per field class, log level, timeouts…). Applies to every key within 30 seconds.
keys_options_set	PUT /v1/keys/{id}/options	keys:write	Default ExecutionOptions for one API key, between the preset and the workspace defaults. Send {} to clear.
cache_stats	GET /v1/cache/stats	logs:read	Records (entities and exact responses), bytes, hits and the vendor spend cache hits saved this month and all time, plus daily hit/partial/miss counts. Cache hits are always free.
cache_list	GET /v1/cache	logs:read	The cache explorer: each cached entity or response with its masked label, fields, their ages and freshness, hits and savings. Values are never returned here.
cache_purge	DELETE /v1/cache	cache:write	Delete cached records by key, tag or capability, or everything with all=true (crypto-shreds the workspace cache key: nothing sealed before can be read again).
cache_forget	POST /v1/cache/forget	cache:write	DSAR delete for the private cache: every cached record about this subject (entity records under all its identities and exact responses naming it) is removed.
cache_feedback	POST /v1/cache/feedback	cache:write	Negative feedback, such as a bounce: the named fields are dropped from the subject's cached record so the next request fetches them again.
logs_tail	GET /v1/logs/live/recent	logs:read	The newest events from the live tail (calls, waterfall run steps, jobs, lists and alerts), newest last, with the same filters as the stream: provider, capability, status, key, app, run or waterfall, minimum latency and cost, sample and fields. Stream them with GET /v1/logs/live or the logs://live MCP resource.
logs_live_ticket	POST /v1/logs/live/ticket	logs:read	A single-use ticket valid for 60 seconds that opens GET /v1/logs/live without an Authorization header (browsers). Carries logs:read for this org only.
drains_list	GET /v1/drains	logs:read	Log drains with their target, filter, batching and delivery health. Credentials are never returned.
drains_create	POST /v1/drains	credentials:write	Send the live log to an HTTPS webhook (HMAC-signed), Axiom, Datadog, or S3, GCS or R2 as NDJSON. Credentials are stored encrypted in the vault and are write-only. Webhook drains return their signing secret once.
drains_get	GET /v1/drains/{id}	logs:read	One drain with its delivery health.
drains_update	PUT /v1/drains/{id}	credentials:write	Rename, pause or resume, change the target, filter or batching, or replace credentials.
drains_delete	DELETE /v1/drains/{id}	credentials:write	Stop sending and remove the drain and its stored credentials.
drains_test	POST /v1/drains/{id}/test	credentials:write	Send one sample event to the drain now and report the result; updates drain health.
alerts_summary	GET /v1/alerts/summary	logs:read	Open alerts by severity, when the last one fired, and how many rules, destinations and webhook endpoints the workspace has.
alerts_rules_list	GET /v1/alerts/rules	logs:read	Every alert rule with its metric, threshold, window, scope, severity, destinations, mute and live state (ok, firing, muted, no_data). New workspaces start with defaults (a vendor key failing, budgets at 80% and 100%, billing mismatches, error-rate and spend spikes, a key on a new IP, failing deliveries, security events).
alerts_rules_create	POST /v1/alerts/rules	keys:write	Alert when a metric crosses a threshold over a window: error_rate, p95_latency, spend, budget_used, vendor_failure_spike, vendor_key_failing, billing_mismatch, cache_hit_rate, delivery_backlog, security_event or new_ip. Scope it to the workspace, a key, a vendor or a capability; route it to every destination or to chosen ones.
alerts_rules_get	GET /v1/alerts/rules/{id}	logs:read	One alert rule with its state.
alerts_rules_update	PUT /v1/alerts/rules/{id}	keys:write	Change any field; fields you leave out keep their value. Changing what the rule measures resolves its open alerts.
alerts_rules_delete	DELETE /v1/alerts/rules/{id}	keys:write	Delete the rule and resolve its open alerts.
alerts_rules_mute	POST /v1/alerts/rules/{id}/mute	keys:write	Snooze a rule for minutes (or until until); it keeps evaluating and recording alerts but sends nothing. minutes: 0 unmutes.
alerts_rules_test	POST /v1/alerts/rules/{id}/test	keys:write	Send a test alert to every destination the rule notifies, now, and report each result. Nothing is recorded as a real alert.
alerts_destinations_list	GET /v1/alerts/destinations	logs:read	Where alerts go (email, webhook endpoints, Slack, Discord, Microsoft Teams, PagerDuty, Opsgenie) with delivery health. Secrets are masked.
alerts_destinations_create	POST /v1/alerts/destinations	credentials:write	Connect email recipients (with optional hourly or daily digests), a webhook endpoint, a Slack or Discord incoming webhook, a Microsoft Teams Workflows URL, a PagerDuty Events API v2 integration key or an Opsgenie API key. The secret is stored encrypted in the vault and never returned.
alerts_destinations_get	GET /v1/alerts/destinations/{id}	logs:read	One destination with its delivery health.
alerts_destinations_update	PUT /v1/alerts/destinations/{id}	credentials:write	Rename, pause or resume, change the minimum severity or settings, or replace the secret.
alerts_destinations_delete	DELETE /v1/alerts/destinations/{id}	credentials:write	Stop sending there and delete its stored secret. Rules that listed it stop routing to it.
alerts_destinations_test	POST /v1/alerts/destinations/{id}/test	credentials:write	Deliver a test alert now and report the result (PagerDuty and Opsgenie tests open and immediately resolve their own incident). Test emails are queued.
alerts_list	GET /v1/alerts	logs:read	Alerts newest first: open (triggered or acknowledged) or resolved, with how often each repeated and its timeline of notifications.
alerts_get	GET /v1/alerts/{id}	logs:read	One alert with its full timeline.
alerts_acknowledge	POST /v1/alerts/{id}/acknowledge	keys:write	Mark an open alert as being handled. PagerDuty and Opsgenie destinations are acknowledged too.
alerts_resolve	POST /v1/alerts/{id}/resolve	keys:write	Close an open alert now; destinations that get resolve notifications are told. The rule's cooldown starts from here.
webhooks_event_types	GET /v1/webhooks/event-types	logs:read	Every event a webhook endpoint can subscribe to, with when it is sent and the current api_version.
webhooks_endpoints_list	GET /v1/webhooks/endpoints	logs:read	Webhook endpoints with their subscriptions, status and delivery health. Secrets are never returned.
webhooks_endpoints_create	POST /v1/webhooks/endpoints	credentials:write	Receive events at an HTTPS URL, signed per Standard Webhooks (webhook-id, webhook-timestamp, webhook-signature). Returns the whsec_ signing secret once.
webhooks_endpoints_get	GET /v1/webhooks/endpoints/{id}	logs:read	One endpoint with its delivery health.
webhooks_endpoints_update	PUT /v1/webhooks/endpoints/{id}	credentials:write	Change the URL, description or subscribed events, or enable and disable it (re-enabling an automatically disabled endpoint clears its failure count).
webhooks_endpoints_delete	DELETE /v1/webhooks/endpoints/{id}	credentials:write	Stop delivering, delete its secrets and delivery log.
webhooks_endpoints_rotate_secret	POST /v1/webhooks/endpoints/{id}/rotate-secret	credentials:write	Issue a new signing secret. The old one keeps signing alongside it for overlap_s (default 24 hours), so both verify while you deploy the new one. Returns the new secret once.
webhooks_endpoints_test	POST /v1/webhooks/endpoints/{id}/test	credentials:write	Deliver an example event of the given type (marked test: true) to the endpoint now and return the delivery with its attempt.
webhooks_deliveries_list	GET /v1/webhooks/deliveries	logs:read	Recent deliveries newest first with every attempt's status code, latency and a redacted response snippet. Kept 30 days.
webhooks_deliveries_get	GET /v1/webhooks/deliveries/{id}	logs:read	One delivery with its attempts and the start of the signed body.
webhooks_deliveries_replay	POST /v1/webhooks/deliveries/{id}/replay	credentials:write	Send a delivery again with the same webhook-id (receivers that dedupe on it ignore a duplicate) and a fresh retry schedule.