Skip to content
Agentmart
Submit a tool

Lastest

Run visual regression tests, review diffs, and manage baselines on a Lastest instance.

Connector by las-team · GitHub account verified by the MCP Registry

IndexedLicense unknownLocal (stdio)★ 21
Write a review

Install

Works with Claude Code, Codex, VS Code, Cursor, Windsurf, Claude Desktop. Read the repository before you give any tool access to your data.

npx agentmarket install connector/las-team-lastest

Asks which of your tools to add it to (Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, Claude Desktop) and shows the change first. See https://agentmart.dev/cli.

Get the weekly diff

The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.

We use your email only for this newsletter. Privacy

About

Run visual regression tests, review diffs, and manage baselines on a Lastest instance.

Reviews

Write a review

No reviews yet. Installed it? Reviews come from verified installers only. How reviews work

Permissions

What this server can do and what it asks for, from its tool list and install config. Checked Oct 25, 2026.

Tools
9 read · 9 write · 0 delete · 11 not yet classified
Secrets
None asked for
Install scripts
Not scanned yet
Runs as
Node.js package

Tools by what they do 29

Read from the source code; the running server may list different tools.

Read 9 · Look things up without changing anything.

  • lastest_explorer_findings
  • lastest_explorer_status
  • lastest_get_diffs
  • lastest_quickstart_status
  • lastest_ranger_status
  • lastest_status
  • lastest_suggest_app_fix
  • lastest_validate_diff
  • lastest_verify

Write 9 · Create, change, send or run things.

  • lastest_approve_layer
  • lastest_build
  • lastest_create_test
  • lastest_explorer
  • lastest_heal_test
  • lastest_publish_share
  • lastest_ranger
  • lastest_run_tests
  • lastest_share

Not yet classified 11 · We could not tell from the name and description.

  • lastest_area
  • lastest_decide_diff
  • lastest_explorer_learn
  • lastest_insights
  • lastest_qa_agent
  • lastest_quickstart
  • lastest_repo
  • lastest_scout_url
  • lastest_setup_script
  • lastest_storage_state
  • lastest_test

Tools (29)

Read from the source code; the running server may differ. Descriptions are the server’s own words; hidden characters are shown as [U+…].

  • lastest_approve_layer

    Per-layer feedback on a step comparison: approve (Mark expected → write baseline), reject (Needs fix → create todo), or snooze (suppress for this build only).

  • lastest_area

    Functional-area resource (test groupings). `action`: "list" (areas for a repo — needs repositoryId), "create" (new area — needs name; optional repositoryId/parentId), "update" (rename/describe/reparent — needs functionalAreaId), "delete" (soft-delete; tests become unassigned — needs functionalAreaId), "list_tests" (tests within an area — needs functionalAreaId).

  • lastest_build

    Build resource. `action`: "list" (recent builds for a repo with status/test counts — needs repositoryId; optional limit), "get" (current status & results of a build; slim diff index by default — needs buildId; pass includeDiffs:"full" for joined a11y/network/AI payloads), "review" (comprehensive QA review: build details + visual diffs + failed tests into a structured summary with action items — needs buildId).

  • lastest_create_test

    Create a test. **Preferred (MCP-first): DIRECT browser mode** — YOU write the Playwright code and pass { name, code }. First read the resource lastest://repo/{repositoryId}/authoring-guide for the exact runner contract + this repo's base URL, areas, and setup; discover selectors with your Playwright MCP (or lastest_scout_url). The `author-test` prompt walks the whole flow. Other modes: **AI browser** ({ url } and/or { prompt }) has the Lastest AI generate the test server-side (only if in-product AI is configured there — else fall back to direct). **direct API** (E1) — { name, testType:"api", a…

  • lastest_decide_diff

    Approve or reject visual diffs (updates baselines / marks regressions). `action`: "approve" or "reject". Provide either `diffIds` (a batch of one or more diff IDs to approve/reject) OR, for action:"approve", a `buildId` to approve ALL pending diffs in that build at once. Exactly one of diffIds / buildId is required.

  • lastest_explorer

    Start the Explorer — an autonomous exploratory-testing agent (explorbot-style). It loops research → plan → act → analyze on a live Embedded Browser: maps each page, drafts scenarios in rotating styles (normal/curious/psycho), drives the browser through them adapting as it goes, records defect/UX findings clustered by root cause, learns per-page experience reused on later runs, and keeps passing flows as quarantined tests. WATCHABLE live in the Lastest activity feed. Async: returns a sessionId — poll lastest_explorer_status.

  • lastest_explorer_findings

    List the full finding rows (defects + UX issues with severity, root-cause cluster, page state, and evidence) recorded by an explorer session.

  • lastest_explorer_learn

    Teach the Explorer about the app: store a knowledge note (markdown hint) matched to pages by URL pattern — quirks, form rules, test data, navigation notes. Matching notes are injected into the explorer's planning and testing prompts on later runs (explorbot's /learn).

  • lastest_explorer_status

    Poll an explorer session: status, current iteration, live stream URL, findings summary, and — once completed — the root-cause-clustered report plus any kept test ids.

  • lastest_get_diffs

    Read visual diffs. `scope`: "single" (full details of one visual diff incl. pixel data, AI analysis, test info — needs diffId), "build" (all visual diffs for a build with AI classification/confidence + aiAnalysis commentary — needs buildId). For "build", pass `full: true` only if you also need the heavy joined payloads.

  • lastest_heal_test

    Trigger the AI healer agent to automatically fix a failing test by inspecting the live UI and updating selectors/assertions.

  • lastest_insights

    Repository-level insights (needs repositoryId). `action`: "coverage" (STRUCTURAL coverage — which routes and functional areas have any test at all), "data_coverage" (DATA coverage — the combinatorial cell model: cell/t-way/weighted-volume coverage, the stopping decision, the highest-weight uncovered cells, and the coverage trend over recent builds), "qa" (comprehensive QA overview: test health, recent builds, and action items). Use "data_coverage" for "are we testing enough of the data space / is coverage improving"; "coverage" only answers "which pages have a test".

  • lastest_publish_share

    Publish a public-share link for a build (or a single test within it). Returns a `/r/<slug>` URL anyone can view without logging in. Use after a build completes so demos and outreach messages can link directly to the visual result. Pass `scopedTestId` to scope the share to one test instead of the whole build.

  • lastest_qa_agent

    Drive the ongoing QA agent for a repository (needs repositoryId). `action`: "status" (live session, latest coverage summary, task board, trigger config), "start_run" (start an autonomous run — optional mode full|refresh_spec|fill_gaps and targetUrl; 409 when a session is already running), "run_status" (poll a session by sessionId), "add_task" (queue a directive on the agent's task board — it picks tasks up when idle, works them with the right protocol, and replies on the card), "list_tasks" (the task board).

  • lastest_quickstart

    Productized form of /gtm-lastest-saas-demo. Spins up a 2-test demo (auth setup + app walkthrough) on a repo whose baseUrl is set, runs it with video, and writes build_demo_notes. Gated by team early-adopter mode + repo baseUrl. Returns a sessionId to poll with lastest_quickstart_status.

Show 14 more tools
  • lastest_quickstart_status

    Poll a QuickStart session by ID. Returns step-by-step status, the auth-setup outcome, the walkthrough test ID, the build ID, and whether demo notes were written.

  • lastest_ranger

    Start a 'ranger' — an Embedded Browser that navigates a URL live and returns a rendered (SPA-aware) page map: headings, landmarks, forms, inputs, buttons, links, test-ids, and candidate selectors. Unlike lastest_scout_url (static, instant), ranger drives a real browser and is WATCHABLE in the Lastest activity feed via a live stream. Async: returns a sessionId immediately — poll lastest_ranger_status for the live streamUrl and the final page map. Prefer your own Playwright MCP if you have one; use ranger when you don't, or want a watchable run on a JS-rendered page.

  • lastest_ranger_status

    Poll a ranger session: returns its status, the live stream URL (watchable while it runs), and — once completed — the rendered page map to author tests from.

  • lastest_repo

    Repository resource. `action`: "list" (all repos for the team), "get" (one repo — needs repositoryId), "create" (new local repo — needs name; optional baseUrl points the repo at an external app so generated tests target the right origin), "update" (rename / branches / baseUrl — needs repositoryId), "get_settings" (repo-level Playwright settings — needs repositoryId), "update_settings" (upsert repo-level Playwright settings — needs repositoryId + at least one field).

  • lastest_run_tests

    Trigger a test build and return structured results. Can run all tests in a repo, all tests inside one functional area, or specific test IDs. Pass `forceVideoRecording: true` when you intend to publish a share that should include the video player (the share page renders the embedded clip only when the build has video).

  • lastest_scout_url

    Static (no-browser) scout of a URL: returns title, headings, forms, inputs, links, and candidate selectors to help you author a test. Best-effort — SPA/JS-rendered content won't appear, so prefer your own Playwright MCP for live pages and use this only as a fallback or quick map.

  • lastest_setup_script

    Setup-script resource — reusable Playwright/API setup blocks. `action`: "list" (scripts for a repo with id/name/type/code — needs repositoryId), "get" (one script incl. code — needs setupScriptId), "create" (new script — needs repositoryId + name + type + code), "update" (name/type/code/description — needs setupScriptId), "delete" (remove; 409 if a test still references it — needs setupScriptId). Attach a script to a test via lastest_test action:"update" with setupScriptId.

  • lastest_share

    Manage existing public shares (publishing is lastest_publish_share). `action`: "list" (shares anchored on a build — needs buildId — or on a test — needs testId; includes revoked ones, check `status`), "revoke" (kill a share by its share ID so the /r/<slug> URL stops resolving — needs shareId; find share IDs via a "list" call).

  • lastest_status

    Instance & background-job status. `action`: "health" (check connectivity to the Lastest instance), "jobs" (list currently active background jobs — builds, AI operations, etc.), "job" (get status/progress of a specific background job — requires `jobId`).

  • lastest_storage_state

    Storage-state resource — saved Playwright `storageState()` blobs (cookies + localStorage). `action`: "list" (metadata only for a repo; raw JSON omitted because it holds live auth tokens — needs repositoryId), "create" (save a new state from a storageState() JSON string — needs repositoryId + name + storageStateJson), "delete" (remove a state — needs storageStateId). Wire a state into a test with lastest_test action:"update" using setupOverrides.extraSteps.

  • lastest_suggest_app_fix

    For a failing test classified as a real regression, get a structured APPLICATION-code fix recommendation (file, snippet, rationale) localized against the build's change map. This is the "fix the app" loop: it complements lastest_heal_test (which fixes the *test*). The suggestion is advisory only — Lastest never edits your application code; review and apply it yourself. Returns `not_a_regression` when the failure was triaged as flaky/environment/test-maintenance.

  • lastest_test

    Test resource (read/update/delete). `action`: "list" (all tests in a repo with pass/fail status — needs repositoryId; pass filter:"failing" to return only failing tests with error details), "get" (full details of one test incl. code/URL/last run — needs testId), "update" (name/code/URL/area/setup wiring/runtime overrides — needs testId), "delete" (soft-delete, restorable — needs testId). To create a test use lastest_create_test; to auto-fix a failing test use lastest_heal_test. For update, pass `setupTestId` to point this test at another test for setup, or `setupScriptId` to point at a saved s…

  • lastest_validate_diff

    Diff-scoped validation in one call: pass a git diff (or a base/head branch range for GitHub repos) and Lastest maps the changed files to the affected tests, runs ONLY those, and returns a pass/fail/review verdict with the failing tests and pending visual changes. Use this in a coding-agent loop right after making a change to confirm nothing relevant broke, without running the whole suite. Blocks until the scoped build finishes by default; pass `wait: false` to get a buildId to poll instead.

  • lastest_verify

    Verify-phase reads for a build (needs buildId). `action`: "view" (full verify-build view — Change Map + step comparisons grouped by regression vs intent gate, plus `visualUrlsByDiffId` clickable /api/media URLs and `testsByTestId` source/setup hints), "change_map" (just the build-level Change Map — 4-signal area ranking + AI intent/risk summary).

Signals

Stars
21
Forks
0
Open issues
0
Contributors
4
Weekly downloads
31
Last commit (this folder)
Sep 1, 2026 (1 month ago)
Last release
Jun 15, 2026 (4 months ago)
Last package publish
May 14, 2026 (5 months ago)

From GitHub (fetched Oct 24, 2026) and the package registry.

Details

Registry name
io.github.las-team/lastest
Version
0.3.7
Auth
No auth
Packages
npm @lastest/mcp-server
License
License unknown
Repository
github.com/las-team/lastest
Website
github.com/las-team/lastest/wiki/MCP-Server
First seen
Oct 24, 2026
Last checked
Oct 24, 2026
Source
Official MCP Registrygithub-graphql
Cite this page

Lastest MCP server. Agent Mart. https://agentmart.dev/connectors/las-team-lastest (last updated 2026-10-24)