addDependency
ChatGPTRecords in the infrastructure graph that one resource depends on another, when Polylane did not discover the connection itself, so blast radius, paths and maps include it. Choose the relationship that matches the dependency. Recording the same connection again replaces its stored properties. Use removeDependency to delete a connection.
addIssueNote
ChatGPTAdds a note or a milestone to an issue's timeline, for example what the team found or decided. It adds an entry and changes nothing else on the issue. Use editIssueNote to change a note that already exists.
archiveFix
ChatGPTArchives one fix so it leaves every fix list and keeps its history. If its pull request is still open, Polylane closes it on GitHub with a comment first, retargeting open pull requests built on its branch to its base branch. Use it to clear fixes the team is done with.
archiveIssue
ChatGPTArchives one issue: it leaves every issue list and keeps its history. Use it to clear an issue the team no longer needs to see, after confirming with the user. To change an issue's status or severity instead, use updateIssue.
cancelFix
ChatGPTStops one fix that is still running, before it opens a pull request, and discards its unfinished work. A fix running on a hosted code agent is stopped there too when its integration is still connected: its Devin session is cancelled, its Factory computer is deleted, its Conductor workspace is archived, or its Capy thread is interrupted and archived. Cursor has no way to stop a running agent, so a Cursor fix may keep working there while Polylane stops tracking it and discards its result. Use closeFix instead once the fix has a pull request open.
closeFix
ChatGPTCloses the open pull request of one fix on GitHub without merging it, and posts the user's comment on it saying why. Open pull requests built on its branch are retargeted to its base branch first. GitHub notifies the pull requests' watchers. Use cancelFix for a fix that has not opened a pull request yet, and recordFixCloseReason to record why a pull request that is already closed was rejected.
createTicket
ChatGPTFiles a new issue in the connected Linear workspace with a title, a markdown description and optionally a state, priority, assignee, project, labels and due date. The team sees it in Linear at once, so confirm the content with the user first. Find the team, state and assignee ids with listTicketTeams.
deleteMemory
ChatGPTDeletes one markdown file from the workspace memory and commits it. Read the file first with readMemory and pass its sha as expectedSha, so a change made in between is not deleted. The memory index MEMORY.md cannot be deleted; edit it with writeMemory instead. Confirm with the user first: the file is gone from the current memory, though the repository history keeps it.
dismissRepositoryAdvisory
ChatGPTIgnores one observability finding for one connected repository, or restores it with dismissed false. Find the finding id with getInstrumentation, which lists each repository finding with its id. For a finding on a cloud resource, use dismissResourceAdvisory.
dismissResourceAdvisory
ChatGPTIgnores one best-practice finding for one cloud resource everywhere Polylane shows it, or restores it with dismissed false. Find the advisoryId and the resource with listAdvisories. For a finding on a repository, use dismissRepositoryAdvisory.
editIssueNote
ChatGPTChanges a note or milestone a person added to an issue's timeline: its text, title, entry type or time. Entries Polylane recorded itself cannot be edited. Find the note id with getIssueTimeline.
escalate
ChatGPTAdds a new escalation: something a person needs to act on, with a short headline and the detail they need. Link it to an issue or an investigation when it comes from one; linked to an issue, it also appears on that issue's timeline and takes the issue's resource. Name the resource when it is about a different one, or when no issue is linked, so people can filter escalations by resource. It appears in listEscalations and sends no message. Check listEscalations first so the same request is not raised twice.
explainFeatureFlagEvaluation
ChatGPTExplains which value one person gets for PostHog feature flags and why: the condition each flag matched for that person's distinct id. Use it when a user asks why someone sees, or does not see, a feature.
explainResource
ChatGPTExplains one resource in plain terms: what it is, its alternate names, its tier, its environment, who owns it, which repository deploys it, and the resources it connects to within two hops, grouped by relationship. Use it first when the user asks what something is or how it fits in. For the raw configuration use getResource; for every edge with its properties use getConnections.
findDeployingRepository
ChatGPTFinds the code repositories that deploy one resource, as Polylane linked them from deploy configuration, infrastructure as code, tags and container images. Use it to go from a running resource to its code, then listCommits or listPullRequests in that repository. An empty result means no repository link is known.
findImageUsage
ChatGPTFinds the resources that run a container image, and the registry that holds it, from the image references Polylane read on running resources. Name the image repository as it appears in an image reference without the tag, for example ghcr.io/acme/checkout, or pass its resource id. Use findDeployingRepository on a result to reach the code that builds it.
findOwner
ChatGPTFinds the people and teams that own one resource, as Polylane learned them from CODEOWNERS files, the provider's creator and modifier records, and tags. Use it to answer who to ask about a resource. An empty result means no owner is recorded; setResourceOwner records one.
findPath
ChatGPTFinds the shortest chain of connections from one resource to another, for questions such as how the website reaches this database. Returns the resources along the path in order. found false means no path exists within maxHops in that direction, not that a resource is missing. Use mapService to see a whole neighbourhood instead.
findResources
ChatGPTSearches the workspace's infrastructure graph: cloud resources across every connected provider, repositories, people and teams, container images and external services. Search by part of a name, title, class, alternate name or id. Without a query it lists resources by type, provider, account, region or category. Every result carries the id the other resource tools take. Only connected accounts are in the graph, so an empty result can mean the account is not connected or still syncing.
findSinglePointsOfFailure
ChatGPTRanks the workspace's critical resources by how much of the infrastructure a failure at each one would reach, so the user sees which single resources carry the most risk. It checks up to 25 tier 1 resources and returns the top ones with how many resources each would affect and a sample of them. Use getBlastRadius to see the full reach of one of them.
fixPullRequestReview
ChatGPTStarts a fix for the problems Polylane's review found in one pull request, which can open a pull request on GitHub with the fix. It links the new fix to the review, replacing any fix linked before. Find the review id (changeId) with listPullRequestReviews and read it with getPullRequestReview. To run an existing fix again, use retryFix.
getAuditLog
ChatGPTReads the audit log of one connected cloud account (AWS, Cloudflare, Vercel or Render): who changed what and when, such as IAM edits, deploys and configuration changes. Use it to find the person or deploy behind a change. Filter by actor, action, resource or status. The last 24 hours when no window is given. AWS needs a region unless account is a Polylane account id.
getBlastRadius
ChatGPTAnswers what breaks if one resource fails: every resource a fault there can reach, with the number of hops and the relationship it travels through. It follows only connections that carry faults, in the direction each one carries them, so it is not the same as listing dependents. truncated true means the radius may reach further than listed. For several resources at once, or a pull request, use getChangeImpact.
getChangeImpact
ChatGPTSizes the impact of changing or losing several resources at once: the combined set of resources a fault at any of them can reach. Name the resources directly, or pass a repository and changeId to start from the resources an assessed pull request affects (find changeId with listPullRequestReviews). Up to 10 starting resources. The result reports unassessed starting resources separately from the graph result's limits. For one resource, getBlastRadius is simpler.
getCommitDiff
ChatGPTReturns what one commit changed: its message, author and date, and the files with the lines added and removed. Find the SHA with listCommits. For a whole pull request's diff, use listPullRequests with includeDiff.
getConnections
ChatGPTLists every direct connection of one resource with the details stored on it: relationship, direction, how it was discovered, and properties such as permissions, ports, bindings or environment variable names. Use it when the user needs how two resources are connected, not only that they are. Narrow by relationship and direction.
getDatabaseLatencyAnomalies
ChatGPTLists the windows in which query latency on one PlanetScale database branch exceeded its baseline, with how long each lasted and the queries most correlated with it. Use getDatabaseQueryInsights for the full query list.
getDatabaseQueryErrors
ChatGPTLists the query errors of one PlanetScale database branch over a time range, grouped by error, with how often each occurred and when it last did. Use it to match an application error window to database failures. Use getDatabaseQueryInsights for the queries themselves.
getDatabaseQueryInsights
ChatGPTLists the query patterns of one PlanetScale database branch over a time range with their counts, errors, latency percentiles and rows read, so the user sees which queries are slow or hot. Use getDatabaseQueryErrors for failing queries and getDatabaseLatencyAnomalies for windows of high latency.
getFeatureFlag
ChatGPTReturns one PostHog feature flag in full in one answer: its definition and rollout conditions, its status, and the flags that depend on it. Use listFeatureFlagChanges for who changed it and when, and explainFeatureFlagEvaluation for why one person gets a value. A part that cannot be read comes back empty with a note.
getFixDiff
ChatGPTReturns the unified diff of the pull request one fix opened, so the user can review the exact code change before merging it with mergeFix or asking for changes with requestFixChanges.
getInstrumentation
ChatGPTReports how well one repository's code is instrumented for observability: which routes emit telemetry and the open findings about what is missing. Use it to answer whether a service's code is observable before relying on its logs or traces.
getInventory
ChatGPTCounts the resources in the infrastructure graph, in total or grouped by provider, account, region, type, tier or category, for questions such as how many Lambda functions there are or what runs in each region. Narrow by provider, account, region and type. A category or environment filter counts the matching total and cannot be combined with groupBy. Use findResources to list the resources themselves.
getInvestigation
ChatGPTReads one investigation: its name, state and the latest messages, including Polylane's findings. While the state is running, Polylane is still working; call again later for the answer. Pass messageId and sourceId to also see an item that was attached to a message, as Polylane read it.
getIssue
ChatGPTReturns one issue in full: what fired, where, its metrics and logs for an occurrence, its analysis, the duplicate and related issues Polylane linked to it, and the action decisions recorded on it. Use it after listIssues when the user asks about one issue. For the event history of the issue, use getIssueTimeline.
getIssueStats
ChatGPTCounts the issues detected and resolved in the workspace over a time window, for questions such as how many issues there were this week. Returns totals only; use listIssues to see the issues themselves.
getIssueTimeline
ChatGPTLists what happened on one issue in order: detection, occurrences, status and severity changes, investigation steps, fixes and notes people added. Use it to explain how an issue developed or what was already done about it. Narrow by a time window or by event types.
getJobQueues
ChatGPTLists the queues of a connected Trigger.dev environment, deepest backlog first, with their running and queued counts, whether they are paused, and their concurrency limit. Pass name (and type) for one queue in full together with its metrics over a period: wait time percentiles, throughput, throttling and how the backlog moved. Use listJobRuns for the runs themselves.
getLogPatterns
ChatGPTReturns the recurring log patterns Polylane has learned for one resource across its monitoring runs, each with how often it appears and how that rate has moved, so the user can see which messages are new or growing. Use queryLogs for the lines themselves.
getMetrics
ChatGPTReturns the key metrics of one resource from its provider over a window, each compared with the window of the same length just before it, such as requests, errors, latency or CPU depending on the resource. The last 24 hours when no window is given. Use getResourceHealth to see the metrics together with the resource's open issues.
getPullRequestReview
ChatGPTReturns one pull request impact assessment Polylane wrote in full: the verdict, the summary, the failure modes it checked and the resources the change affects. Find the assessment id (changeId) with listPullRequestReviews. Use getChangeImpact to size what those resources reach, and fixPullRequestReview to fix what it found.
getResource
ChatGPTReturns one infrastructure resource's full record: its identity, tier, alternate names, timestamps and the configuration Polylane synced from the provider. Use it when the user asks how something is configured. For a plain summary of what the resource is and what it connects to, use explainResource.
getResourceHealth
ChatGPTShows how one resource is doing: its key metrics over the window compared with the window before, and the unresolved issues Polylane has open on it. Use it when the user asks whether something is healthy or what is wrong with it. The last 24 hours when no window is given. Resources without provider metrics return their issues only.
getTopologyPage
ChatGPTReads one page of the workspace's infrastructure graph for the infrastructure map: up to 150 resources and their connections, with a cursor for the next page. The map calls it while it draws the workspace; use findResources or mapService in the conversation.
getWorkspaceOverview
ChatGPTGives a first look at a workspace in one answer: how many issues were detected and resolved over a period, and how many resources the infrastructure graph holds by provider. Use it to orient before narrowing down with listIssues or findResources. The last 7 days when no window is given. A part that cannot be read comes back empty with a note.
investigateIssue
ChatGPTAsks Polylane to investigate one issue and work on a fix: it finds the cause and, when the fix is in code, opens a pull request on GitHub. An issue has one fix run thread: if it exists, the request goes to it, whether its last run is live or ended. Follow progress with getInvestigation and the fix with listFixes.
listAdvisories
ChatGPTLists the best-practice findings Polylane has for the workspace's resources and repositories, grouped by rule: observability, resilience, security, data protection and lifecycle, warnings first, with the affected resources and how to fix them. Filter by category, severity, or whether the finding is on a resource or a repository. Up to 20 resources are shown per finding, with the full count. Use dismissResourceAdvisory or dismissRepositoryAdvisory to ignore a finding.
listAxiomDatasetFields
ChatGPTLists the fields of one Axiom dataset with their types and units, so a runAxiomQuery or runAxiomMetricsQuery query uses names that exist.
listAxiomDatasets
ChatGPTLists the datasets of a connected Axiom organization with their kind and retention, so a query names the right one. Use listAxiomDatasetFields for the fields of one dataset.
listChecks
ChatGPTShows Polylane's anomaly checks. Pass resource for that resource's recent check runs, newest first, and add id for one run in full with its evaluation. Without resource it counts the workspace's check runs over a window: total, healthy, stable and breached. The last 7 days when no window is given.
listCiRuns
ChatGPTLists the recent GitHub Actions runs of one repository with their status, result, branch, commit and timing, to answer whether the build passed. Filter by branch, triggering event or status. Pass runId to read one run in full. Job logs are not included.
listCloudflareTelemetryKeys
ChatGPTLists the field names present in a connected Cloudflare account's Workers and Containers telemetry, optionally matching a substring, so a query can filter or group by the right key. Use it before runCloudflareTelemetryQuery or searchCloudflareTelemetryEvents, and listCloudflareTelemetryValues for a field's values.
listCloudflareTelemetryValues
ChatGPTLists the distinct values one field takes in a connected Cloudflare account's Workers and Containers telemetry, such as the script names or outcomes seen, so a filter uses real values. Find the field with listCloudflareTelemetryKeys first.
listCommits
ChatGPTLists the commit history of one connected repository, newest first, with each commit's SHA, message, author and date. Filter by branch, file path, author and dates. Use getCommitDiff for the lines one commit changed, and listPullRequests for open pull requests.
listCriticalResources
ChatGPTLists the resources Polylane ranks as critical (tier 1): the ones the architecture depends on and that Polylane watches most closely. Set includeStandard to add tier 2 resources. Narrow by provider or type. Use it to answer what matters most in this workspace; use findSinglePointsOfFailure to rank them by how much breaks when they fail.
listDependencies
ChatGPTLists the resources one resource uses directly: what it calls, reads, publishes to or runs on, one hop away. Narrow by relationship or by the type of the other resource. Use listDependents for the opposite direction, findPath to connect two resources, and mapService for more than one hop.
listDependents
ChatGPTLists the resources that use one resource directly: its callers, consumers, owners and the repositories that deploy it, one hop away. Narrow by relationship or by the type of the other resource. Use listDependencies for the opposite direction and getBlastRadius for what a failure would reach.
listEscalations
ChatGPTLists the escalations in the workspace: things Polylane or a person raised for someone to act on, such as a credential to renew or a decision an investigation could not make, with how often each recurred. Open ones by default; filter by status or search the headline and detail. Pass id to read one in full with every occurrence folded into it.
listExternalDependencies
ChatGPTLists the third-party services the infrastructure depends on, as Polylane found them in configuration, with the resources that use each one. Pass resource to list only the external services that one resource uses. Without it, up to 15 external services are listed with their users.
listFeatureFlagChanges
ChatGPTShows how PostHog feature flags change. Pass id for one flag's change history: who changed it, when, and what. Without id it lists the scheduled changes waiting to apply across the project. On a free PostHog plan the change history is not available and says so.
listFeatureFlags
ChatGPTLists the feature flags of a connected PostHog project with their keys, ids and rollout filters, so the user can find the flag behind a change in behaviour. Use getFeatureFlag for one flag in full and listFeatureFlagChanges for its history.
listFixes
ChatGPTLists the fixes Polylane worked on: each fix's status, the issue it addresses, the repository and branch, and its pull request. Filter by status, repository or a search in the title. Pass id to read one fix in full with its timeline of what Polylane did. Use getFixDiff for the code change itself.
listHoneycombColumns
ChatGPTLists the columns of one Honeycomb dataset with their types and descriptions, so a runHoneycombQuery query uses columns that exist.
listHoneycombDatasets
ChatGPTLists the datasets of a connected Honeycomb environment with when each last received events. Use listHoneycombColumns for the columns of one dataset before writing a runHoneycombQuery query.
listInvestigations
ChatGPTLists the workspace's investigations and conversations with Polylane, most recently updated first, with their state: running, waiting on a person, or completed. Pass query to search them by what they are about. Use getInvestigation to read one.
listIssues
ChatGPTLists the issues Polylane tracks in the workspace: alerts, anomalies and defects, newest first, with severity, status and the resource each one is on. Use it to answer what is broken or what fired recently, and to find an issue id for getIssue. Filter by status, severity, source, cause, a time window, a title search, or one resource. Archived issues are left out unless asked for.
listJobRuns
ChatGPTLists the background job runs of a connected Trigger.dev project, newest first, filtered by task, status, tag, schedule, version, error or time. Pass runId for one run in full, with its payload and output when asked. Use listWorkflowRuns for Cloudflare Workflows.
listJobSchedules
ChatGPTLists the schedules of a connected Trigger.dev project, optionally only active ones or one task's, with their cron and next run. Pass scheduleId for one schedule in full.
listLogSources
ChatGPTLists where the workspace's logs live in the connected log providers: AWS CloudWatch log groups (up to 50), Axiom datasets and Better Stack sources. With one connection per provider it picks it on its own. With several, that provider's part names the ids to choose from; pass cloudAccountId (and region) for AWS, axiomIntegrationId or betterstackIntegrationId. Providers the workspace has not connected are named in a note. Use it to find the dataset or log group to query, then runAxiomQuery for Axiom, or queryLogs for one resource.
listMonitoringGaps
ChatGPTShows where monitoring is thin: how many resources Polylane can monitor and how many it does, the observability findings that leave resources without useful telemetry, and the resources set to minimal monitoring, whose changes are recorded but not investigated. Use it to answer what is not being watched. Use setMonitoring to change how closely a resource is watched.
listPullRequestReviews
ChatGPTLists the pull requests of one connected repository that Polylane assessed for production risk, newest first, with each assessment's verdict, impact level and the resources the change affects. Use getPullRequestReview for one assessment in full, and listPullRequests for the pull requests themselves.
listPullRequests
ChatGPTLists the pull requests of one GitHub repository, open ones by default, filtered by state, base or head branch. Pass pullNumber to read one pull request with its reviews and mergeable state, and set includeDiff to add its full diff.
listReports
ChatGPTFinds the written output of Polylane's work. Without inputs it lists the pages Polylane published for the workspace; pass pageId for a short-lived link to view one. Pass investigationId to list the reports and other artifacts one investigation produced, and add reportId to read one in full.
listRepositories
ChatGPTLists the code repositories connected to the workspace, with their sync status. Pass query to search them by name or description, or repository to get one in full. Use the owner/name it returns with the other code tools.
listRepositoryResources
ChatGPTLists the running resources one code repository deploys to, such as its functions, services and sites, so the user can see where a repository's code runs and what a change to it can touch. Name the repository as owner/name or pass its resource id. Use getChangeImpact to size what a change to those resources could break.
listTicketTeams
ChatGPTLists the teams of the connected Linear workspace with their ids. Pass teamId for that team's workflow states and members in one answer, which createTicket and updateTicket need for stateId and assigneeId.
listTickets
ChatGPTReads issues in the connected Linear workspace. Pass issueId for one issue in full, or term to search by text; otherwise it lists issues filtered by team, project, assignee and workflow state. Use createTicket to file a new one.
listTriageRules
ChatGPTShows how Polylane triages incoming alerts: the alert signatures it suppresses, the signatures it tracks as flaky, and, for one connection, which of its alert rules Polylane triages automatically. Use it before updateTriageRule or suppressAlertSignature, or to explain why an alert did or did not become an issue.
listWorkflowRuns
ChatGPTLists the instances of one Cloudflare Workflow in a connected account, with their status, filtered by status and dates; pass instanceId for one instance. Use listJobRuns for Trigger.dev.
listWorkspaces
ChatGPTLists the Polylane workspaces this connection can read, with each workspace's id, slug and name. Call it first when the user has more than one workspace, then pass the id or slug as workspaceId to every other tool.
mapService
ChatGPTDraws the part of the infrastructure graph around one resource: every resource within depth hops (2 by default, at most 3) and the connections between them, as a list and as a Mermaid flowchart the user can render. Use it to show how a service is put together. At most 40 resources are mapped; the result says when the neighbourhood was larger. For one hop with details use getConnections.
mergeFix
ChatGPTMerges the open pull request of one fix into its base branch on GitHub. It cannot be undone from Polylane and the change ships wherever that branch deploys, so read the diff with getFixDiff and get the user's explicit go-ahead first.
muteCloudWatchAlarm
ChatGPTTurns off the actions of one CloudWatch alarm in a connected AWS account, so it stops sending notifications and running its automated actions. The alarm keeps evaluating and its state keeps changing; its actions stay off until unmuteCloudWatchAlarms turns them back on.
openIncidentPanel
ChatGPTOpens the active issues panel beside the conversation: the workspace's active issues, newest first, without dismissed issues or duplicates. Selecting one shows its summary, timeline, investigation and affected resources, and adds it to the conversation's context. Up to 50 issues are listed. Use listIssues to filter issues in the conversation.
openTopologyMap
ChatGPTOpens the workspace's infrastructure map: every resource as a point on a rotating sphere, coloured by category, with rings on resources that have open issues, to search by name or type. Selecting a resource shows its summary, health and open issues, and can open its neighbourhood. The map draws up to 8,000 resources. Use mapService to draw one resource's neighbourhood in the conversation.
queryLogs
ChatGPTReads the recent log lines of one resource from its provider, newest window first, optionally only lines containing some text. It covers Cloudflare Workers, Workers for Platforms, Durable Objects, Containers, AI Gateway and Access applications; AWS Lambda functions, ECS and App Runner services, RDS instances and clusters, EC2 instances, EKS clusters and their Kubernetes workloads; and Vercel, Fly, Render, Supabase, Railway and Trigger.dev resources. For Fly it returns HTTP response rates by status instead of log lines. Other resources return no lines. The search, with its case and regular expression options, is applied by the provider for Cloudflare Workers, Durable Objects, Containers and Access, AWS and Trigger.dev, and checked on each returned line for the others, where a page can then hold fewer lines than limit. At most limit lines come back (100 by default). A returned cursor reads the next page, for Cloudflare Workers, Durable Objects, Containers and AI Gateway and for Vercel request logs; other resources return none, so read older lines by moving to back. The last hour when no window is given. Use getLogPatterns for recurring patterns over time, and the provider query tools for exact queries.
readMemory
ChatGPTReads the workspace memory, a set of markdown files Polylane keeps. Pass a folder path, or nothing for the top folder, to list its files; pass a file path ending in .md to read that file. A file answer carries a sha: pass it to writeMemory as expectedSha when you change that file, so a change made in between is not overwritten.
recordFixCloseReason
ChatGPTRecords why the closed pull request of one fix was not merged, in the user's words: the fix was wrong, a person fixed it another way, it was not a real problem, it went stale, or the behaviour is intended. It replaces any reason recorded before, including one Polylane inferred from the closing comment. Use it after closeFix or after someone closed the pull request on GitHub.
removeDependency
ChatGPTDeletes one connection between two resources from the infrastructure graph. It cannot be undone, and a connection Polylane discovered comes back on the next sync if it still exists. Confirm the connection with getConnections first.
replyToInvestigation
ChatGPTSends a message in one investigation as the user, for example a follow-up question or an answer Polylane asked for. Polylane starts working on it at once and the reply appears in getInvestigation.
reportFeedback
ChatGPTSubmit feedback about the Polylane MCP server to the Polylane support team when the user explicitly requests sending a report. Creates a new feedback record containing the user's approved problem description, such as a confusing result, missing capability, or unclear error. Before submission, tell the user that the report includes workspace and session identifiers, client name/version, user agent, connection settings, tool call count and first tool name, and coarse network location. Obtain the user's approval to send the feedback and these diagnostics to Polylane support. Include only the information needed to explain the problem; omit credentials and unrelated personal data. Requires an OAuth or broker-authenticated user session; API-key sessions cannot submit feedback. The recipient is the fixed Polylane support team. Input: { feedback (what went wrong and what you expected instead), workspaceId? (id or slug; required when the session cannot resolve a single workspace) }.
requestFixChanges
ChatGPTAsks Polylane to revise the open pull request of one fix: the request is posted on the fix's run as the user's message and Polylane pushes new commits to the pull request on GitHub. Say what should change in the user's words.
retryFix
ChatGPTRuns one existing fix again: a fix that was skipped or failed starts over under the same fix, on the hosted code agent it ran on (Devin, Cursor, Factory, Conductor or Capy) when it used one, and a fix waiting for approval opens its pull request on GitHub. Its previous status and run state are replaced. Find the fix with listFixes. To fix what Polylane's review of a pull request found, use fixPullRequestReview; to fix an issue, use investigateIssue.
runAxiomMetricsQuery
ChatGPTRuns one read-only MPL query against a metrics dataset in a connected Axiom organization, for numeric time series such as rates, gauges and percentiles. The query starts with a metric identifier. For logs, traces and events, use runAxiomQuery.
runAxiomQuery
ChatGPTRuns one read-only APL query against a connected Axiom organization, for logs, traces or events: raw rows or aggregations. Start the query with the dataset; find datasets and fields with listAxiomDatasets. For metrics datasets, use runAxiomMetricsQuery.
runCheck
ChatGPTRuns Polylane's anomaly check of one resource now instead of waiting for its schedule: the same evaluation of its metrics and logs that runs on its own, which records a check run and can open an issue when it finds a problem. Read the result later with listChecks. To run an OpenStatus uptime monitor, use runUptimeCheck.
runCloudflareGraphql
ChatGPTRuns one read-only GraphQL query against Cloudflare's Analytics API for a connected account, for data the other Cloudflare tools do not cover, such as Worker and Durable Object invocation counts, CPU time or HTTP request analytics. The query must declare $accountTag and filter accounts on it; the account is filled in. Find datasets and fields with searchCloudflareGraphqlSchema.
runCloudflareTelemetryQuery
ChatGPTRuns one aggregate query over a connected Cloudflare account's Workers and Containers logs and traces, OpenTelemetry data, or AI Gateway request logs: counts, sums, averages and percentiles, grouped by any field, filtered and ordered. Use it for questions such as error rate by script or AI Gateway spend by model. Find field names with listCloudflareTelemetryKeys and their values with listCloudflareTelemetryValues. To read the matching log lines themselves, use searchCloudflareTelemetryEvents. The events are only what the code logged, and may be sampled.
runHoneycombQuery
ChatGPTRuns one read-only query against a dataset in a connected Honeycomb environment, with calculations such as COUNT, P99 or HEATMAP, filters, breakdowns and a time range, returning totals and a time series. Use it to investigate latency, error rate or throughput on traces and events. Find datasets and columns with listHoneycombDatasets.
runUptimeCheck
ChatGPTRuns one check of an OpenStatus uptime monitor now instead of waiting for its next scheduled run, for example to verify a fix. It starts the check in OpenStatus and returns before the result is in. To run Polylane's own check of a resource, use runCheck.
searchCloudflareGraphqlSchema
ChatGPTSearches Cloudflare's Analytics GraphQL schema for types, fields, arguments and enum values matching keywords, such as workers or firewall, so runCloudflareGraphql can be given a valid query.
searchCloudflareTelemetryEvents
ChatGPTReturns the individual log and trace events of a connected Cloudflare account's Workers and Containers that match filters or a full-text search, such as every error-level line from one script in the last hour. Use runCloudflareTelemetryQuery to count or aggregate instead, and listCloudflareTelemetryKeys to find field names. The events are only what the code logged, and may be sampled.
searchEverything
ChatGPTSearches several kinds of workspace records at once, by meaning and by keyword: infrastructure resources, repositories, investigations and connected cloud accounts, ranked by relevance. Use it when the user does not say what kind of thing they mean. Issues and code are not part of it: use listIssues for issues, and findResources for a more precise search of one kind.
searchMentions
ChatGPTFinds the resources, issues and investigations whose names match the text the user typed after @ in the composer, so the user can attach one to a message; an empty query lists critical resources, unresolved issues and recent investigations. Returns at most 20 links, each readable as a resource with a short summary.
setAlternateNames
ChatGPTReplaces the full list of other names one resource goes by, such as the service name its logs use or the name the team calls it, so every Polylane tool finds it by those names. Pass the complete list; names left out are removed, and an empty list clears them. Read the current names with explainResource first.
setMonitoring
ChatGPTTurns Polylane's anomaly monitoring of one resource on or off, or sets how closely it is watched: tier 1 critical, 2 standard, 3 low, 4 minimal. resetTier hands the tier back to Polylane's automatic classification. It replaces the resource's current setting, so read it first with explainResource and confirm with the user.
setResourceOwner
ChatGPTRecords that a person or team owns one resource, so findOwner returns them. The owner must already be a person or team in the workspace's graph; find them with findResources using type polylane.person or polylane.team. Recording the same owner again replaces the stored link.
startInvestigation
ChatGPTAsks Polylane to investigate a question in a new conversation of its own: Polylane reads the workspace's connected providers, telemetry, code and graph, and writes up what it finds. Use it for open questions that need many steps, such as why checkout got slower yesterday. It returns the investigation id at once; the work runs in the background, so read the answer later with getInvestigation and add to it with replyToInvestigation. To fix a tracked issue, use investigateIssue.
suppressAlertSignature
ChatGPTSuppresses one alert signature so matching alerts stop becoming issues, or lifts the suppression with suppressed false. Record why with causeClass and explanation; the team sees them. Find the signature on the issue with getIssue and the current suppressions with listTriageRules. To stop triaging a whole connection or alert rule, use updateTriageRule.
syncCloudAccount
ChatGPTStarts a background sync of one connected cloud account, so Polylane's copy of its resources and connections catches up after a deploy or a provisioning change. It runs on its own and can take several minutes; the graph tools show the new state when it finishes. In Polylane's graph, the sync overwrites every configuration field the provider reports, including fields changed with updateResourceConfiguration. A discovered connection it does not see again is removed at once when both resources were refreshed in this sync, and otherwise after a day unseen. Connections read from a resource's configuration (references in a service's environment variables, Cloudflare Worker bindings, Logpush destinations, AWS IAM wildcard access) are replaced on every complete read. A resource the sync no longer lists can be deleted in that same sync when another resource of its type was refreshed in the same account and region, and otherwise after a day unseen; it is deleted once the provider confirms it is gone, or without a check for types Polylane cannot check, together with every connection to it, including ones recorded with addDependency or setResourceOwner. Connections recorded by hand between remaining resources stay. On an AWS account connected through Polylane's CloudFormation stack, the sync can also write to the account: it turns the recorded CloudTrail trail multi-region when the role allows it, and with alarm creation or subscription turned on, the first sync of a monitored region outside the connection's region creates a Polylane SNS topic and subscription there, creates or overwrites Polylane's default CloudWatch alarms, and adds the topic to the actions of the other alarms there.
unmuteCloudWatchAlarms
ChatGPTTurns the actions of up to 100 CloudWatch alarms in a connected AWS account back on, so they notify and run their automated actions again. Use it to undo muteCloudWatchAlarm.
updateEscalation
ChatGPTChanges one escalation: its headline, detail or category; the resources it concerns, with addResources and removeResources; or with cleared true marks it handled so it leaves the open list and new occurrences Polylane records no longer fold into it (cleared false reopens it). Removing a resource corrects the current list only: a later occurrence on that resource adds it again. Read it with listEscalations first.
updateIssue
ChatGPTChanges one issue's status, its severity, or whether it is resolved. It overwrites those fields and leaves the rest of the issue as it is. Read the issue with getIssue first and confirm the change with the user. To take an issue out of every list, use archiveIssue.
updateResourceConfiguration
ChatGPTSets fields of the configuration Polylane records for one resource: each key you send replaces that key's value and the other keys stay. It changes Polylane's record only, not the resource at the provider, and the next sync of the account overwrites the fields the provider reports. Read the resource with getResource first. To change the names the resource is found by, use setAlternateNames.
updateSentryIssue
ChatGPTChanges one issue in the connected Sentry organization: resolves, unresolves or ignores it, sets its substatus, or assigns it to a user or team. The change is made in Sentry, where the team sees it, so confirm it with the user first. Sentry issues that Polylane tracks also appear in listIssues.
updateTicket
ChatGPTChanges one issue in the connected Linear workspace: its title, description, state, priority, assignee, project, labels or due date. Each field you pass replaces its value and fields you leave out stay as they are. To remove the due date, the project or the assignee, name it in clear. Confirm the change with the user first. To add a comment instead, use commentOnTicket.
updateTriageRule
ChatGPTTurns Polylane's automatic triage of incoming alerts on or off for a whole provider connection or cloud account, or for one of its alert rules when ruleId is set. Alerts that are not triaged do not become issues. Check the current settings with listTriageRules first. To silence one recurring alert signature instead, use suppressAlertSignature.
whoCalls
ChatGPTAnswers who is calling one resource: the callers the infrastructure graph recorded, with how and when each was last seen, merged with the operations the provider's telemetry saw hitting the resource in the window. Use it when a shared component such as a database, queue or function is overloaded or erroring and the user needs the source. The last 24 hours when no window is given.
writeMemory
ChatGPTCreates or replaces one markdown file in the workspace memory and commits it. Polylane reads the memory index and the newest daily notes at the start of every conversation, so keep entries short and true. Read the file first with readMemory and pass its sha as expectedSha; pass null to create a file that must not exist yet. The nightly update owns the .dream folder.
