MCP Server · v3.0.0Beta

Testing intelligence for every AI coding assistant.

Give Cursor, Claude, and Copilot real testing context — journeys, coverage, run history, and auto-heal patches — through the Model Context Protocol. One token. Zero per-service secrets.

View on GitHub

Production · mcp.testrelic.ai + npm @testrelic/mcpMCP Auth Tokenstdio + HTTP · AGPL-3.0

tr_user_journeys()95% users · top 10 journeys

Cursor

Claude

Copilot

Codex

Antigravity

User Journeys

Amplitude · 30d

Test Runs

pass · fail · flaky

142 pass3 fail1 flaky

last run · 38s ago

Coverage Map

95 / 95 target

95%

95 / 95 targets

↑ 4 this week

Auto-Heal

unified diff

-page.click('#submit')

+getByRole('button')…

+14 / −3healed

MCP

@testrelic/mcp · v3.0.0

Cursor❯“What's the user impact of PR #412?”→tr_user_journeys()

streamingp50 12mslive tools

Agent

Cursor / Claude

MCP

@testrelic/mcp

Cloud

TestRelic

Real journeys · Runs · Coverage · Healing

Works with any MCP client

Cursor

Claude Desktop

VS Code Copilot Chat

OpenAI MCP bridge

Anthropic MCP

Any MCP client

Custom agent

Sampling-aware

Cursor

Claude Desktop

VS Code Copilot Chat

OpenAI MCP bridge

Anthropic MCP

Any MCP client

Custom agent

Sampling-aware

Cursor

Claude Desktop

VS Code Copilot Chat

OpenAI MCP bridge

Anthropic MCP

Any MCP client

Custom agent

Sampling-aware

Cursor

Claude Desktop

VS Code Copilot Chat

OpenAI MCP bridge

Anthropic MCP

Any MCP client

Custom agent

Sampling-aware

Four intelligence cases

Context your AI agent actually needs.

A compact, composable tool surface engineered for token efficiency — so your agent stays focused on the work, not on hunting for data.

core

Core & connectivity

Repositories, resolved config, health, and recent runs—always-on primitives every other tool builds on.

Workspace · health + repos

tr_healthnominal

acme-webmobile-appdesign-system

#8841passedPlaywright

#8840failedPlaywright

tr_recent_runs · paginated

triage

Run triage & RCA

Diagnose failures, compare runs, audit flaky tests, pull platform RCA, and open Jira—without leaving the agent.

Run diagnostic

tr_diagnose_runtr_flaky_audit

TimeoutError: locator resolved to hidden element

video markers·ClickHouse flake score

tr_create_jiraTRL-902

signals

Production signals

Blend Amplitude user impact with Grafana Loki patterns so fixes are sized against real sessions, not guesswork.

Impact + logs

tr_user_impact

2.4k

affected users

tr_production_signal

ERRPOST /checkout 5xx spike

WRNretry storm · api-cart

tr_affected_sessions · cohort for comms / rollback

devtools

Dev intelligence

Quality trends, alert posture, code search, and cache stats—keep the agent grounded in how the suite behaves over time.

Quality trends

tr_project_trends14d window

Quality gate

pass rate below threshold

tr_active_alerts

Install in under a minute

One server. Every MCP client.

Production: use the hosted MCP endpoint, run @testrelic/mcp@3.0.0 over npx with your TESTRELIC_MCP_TOKEN, or install the Cursor plugin. Optional sandbox tab covers --mock-mode for non-production only.

Install via

Compatible clients

CursorClaude DesktopCopilot ChatOpenAI MCP

~/.cursor/mcp.jsonJSON

Production streamable HTTP endpoint — no local Node checkout. Same URL family as TestRelic’s prod MCP ingress (mcp.testrelic.ai).

1{
2  "mcpServers": {
3    "testrelic": {
4      "url": "https://mcp.testrelic.ai/mcp"
5    }
6  }
7}

CLI: mcp-server-testrelic login · env: TESTRELIC_MCP_TOKEN

Eight capability groups · 35+ tools

Pick only the tools you need.

Every capability maps 1:1 to a tool group. Enable what the task requires — the agent only ever sees that slice of the schema.

Capabilitycore

Always on · repositories, runs, config, health.

tr_describe_repo()

read

Returns a repo's integrations and capabilities. Sourced from the startup bootstrap — zero additional upstream calls.

tr_get_config()

read

Returns the resolved configuration — capabilities, transport, timeouts, cache/output dirs.

tr_health()

read

Reports upstream connectivity, cache state, and whether any circuit breakers are open.

tr_integration_status()

read

Live health check for one integration type in the current org (e.g. 'jira', 'amplitude', 'grafana-loki').

tr_list_repos()

read

Lists repos the authenticated user can see in cloud-platform-app. No upstream fetch per call.

tr_recent_runs()

read

Paginated list of recent runs. Supports filters by repository, framework, status.

Legacy v1 testrelic_* aliases are registered for every tool to preserve backward compatibility.

Live `--caps` flag

Toggle capabilities above and the CLI command below updates in real time. `core` is always on.

shell

$ npx -y @testrelic/mcp@3.0.0 --caps core,coverage,creation,healing,impact

Enabled

corecoveragecreationhealingimpacttriagesignalsdevtools

Capability gating cuts the tool schema by ~4× for AI agents — only the tools you enable appear in the schema.

Ready-made prompts

Chain tools into complete workflows.

Three first-party prompts ship with the server — exposed to every MCP client the moment the server boots.

coveragecreation

create_test_from_gap

End-to-end journey → test flow

From a coverage gap to a runnable test in one prompt.

Picks a top uncovered journey, drafts a plan, synthesises test code, and runs tsc + --list before returning.

journey gapplangeneratedry-run

triagehealing

triage_and_heal

diagnose → RCA → heal → Jira

One chain to take a red run to a merged fix.

Diagnoses the run, pulls platform RCA, proposes a unified-diff patch, and files (or dedupes) a Jira ticket.

failing runRCApatch diffJira ticket

impact

pr_impact_gate

Risk-rank tests for a diff

Turn PR review into a disciplined gate.

Parses the diff, computes the blast-radius on real Amplitude users, and emits a ranked test selection.

PR diffaffected coderisk scoreMUST / SHOULD / OPTIONAL

> 60 % fewer tokens vs plain LLM prompting

A four-layer cache, engineered for tokens.

Every tool call cascades through L1 → L4 with a cache-first path, returns compact markdown plus structured JSON, and pins token baselines in CI so the savings don't drift.

Tool request

tr_coverage_report{project_id}

Response

cache_key · 95% cov

L1LRU (in-process)

lru-cache · 60s TTL

70%

hit rate

L2SQLite (disk)

better-sqlite3 · 1h – 24h TTL

18%

hit rate

L3Vector (HNSW)

hnswlib + BGE-small · semantic match

hit rate

L4Blob store

SHA256 keyed · large payloads

hit rate

3-state reads (full / unchanged / diff) via SimHashper-tool token budget · auto-truncateclient-side sampling · no server-side LLM key

Capability gating

The LLM only sees tool schemas for the caps you enable — cuts the prompt surface by roughly 4×.

3-state reads

Every read tool supports full / unchanged / diff via SimHash — repeat reads of large payloads cost almost nothing.

Per-tool token budget

Automatic truncation with a cache_key pointer back to the full blob so the agent can opt-in when needed.

Client-side sampling

Code synthesis runs on the client's model via MCP sampling — no server-side LLM key is ever required.

Credential flow

Your machine / CI

One TestRelic token—no third-party secrets stored locally.

TestRelic Cloud

Server-side bootstrap

Jirascoped

Amplitudescoped

Grafana Lokiscoped

GitHubscoped

Per-service credentials stay in the cloud database; they never reach the MCP process.

Per-service secrets are locked inside TestRelic's cloud database. The MCP never sees them.

Security model

One token. Zero third-party secrets.

Authentication collapses to a single personal access token from TestRelic. Every integration is brokered by the TestRelic cloud—so your laptop, CI runners, and editor installs never hold a Jira or Amplitude key.

One credential

You authenticate with a single personal access token from TestRelic. Pass it through the CLI, your shell environment, or a local profile file—never a stack of third-party keys on the machine.

Bootstrap, don't store

Jira, Amplitude, Grafana Loki, and GitHub are resolved server-side in the TestRelic cloud. Per-service credentials never live on your laptop or CI runner.

Capability-scoped surface

The server ships as eight capability groups and dozens of tools. You enable only the slices your workflow needs—so agents see a smaller schema, less context noise, and a tighter operational footprint.

Open source

Read every line. Run it offline.

The plugin surface is intentionally tiny — a manifest, an mcp.json, vendored dist/, a logo, and docs. No binaries, no remote-fetched scripts, no third-party credentials. Reviewers can audit the tool surface in sandbox (--mock-mode) before connecting production data with an MCP token.

testrelic-ai/testrelic-mcp-server SECURITY.md

License

AGPL-3.0

Use, modify, and redistribute.

Package

@testrelic/mcp

Binary: mcp-server-testrelic

Monorepo

packages/mcp + extension

Server + VSCode / Cursor host.

Source

testrelic-ai/testrelic-mcp-server

Pre-built dist/ vendored · no remote fetch.

Ship with context

Give your agent the testing context it needs.

Install via npx or your MCP client and explore every tool against the bundled fixtures. Create an MCP token when you're ready to connect real data — or talk to us first.

Create MCP token Talk to us

npx -y @testrelic/mcp@3.0.0 · https://docs.testrelic.ai/mcp/quickstart