@inspectr/mcplab 1.14.3 → 1.16.0

package/dist/cli.js

@@ -5,7 +5,7 @@ import { Command } from 'commander';
 import kleur from 'kleur';
 import { readFileSync, readdirSync, writeFileSync } from 'node:fs';
 import { join, relative, resolve } from 'node:path';
-import { loadConfig, hashConfig, selectScenarios, runAll, renderSummaryMarkdown, expandConfigForAgents } from '@inspectr/mcplab-core';
+import { loadConfig, hashConfig, selectScenarios, runAll, renderSummaryMarkdown, expandConfigForAgents, applyRuntimeServerOverrides } from '@inspectr/mcplab-core';
 import { renderReport } from '@inspectr/mcplab-reporting';
 import { execSync, spawn } from 'node:child_process';
 import { stringify as stringifyYaml, parse } from 'yaml';
@@ -15,7 +15,6 @@ import { migrateSourceConfig } from './migrate-utils.js';
 import { resolveRunOptions, runInteractiveSelection } from './run-interactive.js';
 import { promptAppOptionsInteractive, selectRunDirInteractive } from './interactive-helpers.js';
 import { deriveConfigRelativePath, resolveRunConfigSelection } from './eval-config-files.js';
-import { applySnapshotPolicyToRunResult, buildSnapshotFromRun, compareRunToSnapshot, formatSnapshotComparisonTable, listSnapshots, loadSnapshot, saveSnapshot } from './snapshot.js';
 import { loadOrBuildSearchIndex } from './results/indexer.js';
 import { searchDocs } from './results/search.js';
 import { formatContext, formatRunList, formatSearchHits, listRuns, showRun } from './results/format.js';
@@ -36,14 +35,13 @@ program
     .option('--agents <agents>', 'Comma-separated list of agents to test (runs each scenario with each agent)')
     .option('--agents-all', 'Run all configured agents for the selected scenarios')
     .option('--interactive', 'Prompt for required inputs')
-    .option('--snapshot-eval', 'Apply snapshot eval policy configured in the config')
-    .option('--compare-snapshot <snapshotId>', 'Compare completed run against snapshot id')
     .option('--bail', 'Stop after first failed config when --config points to a folder')
     .option('--run-note <text>', 'Optional note attached to the run metadata (max 500 chars)')
     .option('--runs-dir <path>', 'Directory for run artifacts', 'mcplab/results/evaluation-runs')
-    .option('--snapshots-dir <path>', 'Directory for snapshots', 'mcplab/snapshots')
     .option('--oauth-token <server=token>', 'Pre-obtained OAuth Bearer token for a server (repeatable, format: server-name=token)', (val, acc) => [...acc, val], [])
     .option('--open-browser', 'Open browser to mcplab serve UI when OAuth is required (default: print URL only)')
+    .option('--server-override-all <serverRef[,serverRef...]>', 'Override MCP server refs for all selected scenarios for this run only')
+    .option('--server-override <scenarioId=serverRef[,serverRef...]>', 'Override MCP server refs for one scenario (repeatable, higher priority than --server-override-all)', (val, acc) => [...acc, val], [])
     .action(async (options) => {
     try {
         const hasAgentOverride = Boolean(options.agents) || Boolean(options.agentsAll);
@@ -57,10 +55,11 @@ program
                 promptAgentSelection: needsAgentPrompt,
                 loadConfigForValidation: (path) => {
                     const loaded = loadConfig(path);
-                    const { agents: libraryAgents } = readLibraries(loaded.bundleRoot);
+                    const { agents: libraryAgents, servers: libraryServers } = readLibraries(loaded.bundleRoot);
                     loaded.config = {
                         ...loaded.config,
-                        agents: { ...libraryAgents, ...loaded.config.agents }
+                        agents: { ...libraryAgents, ...loaded.config.agents },
+                        servers: { ...libraryServers, ...loaded.config.servers }
                     };
                     return loaded;
                 }
@@ -78,19 +77,12 @@ program
         const requestedPath = selection.requestedPath;
         const requestedPathIsDirectory = selection.requestedPathIsDirectory;
         const isBatch = requestedPathIsDirectory;
-        if (isBatch && options.compareSnapshot) {
-            throw new Error('--compare-snapshot is not supported when running a config folder');
-        }
         if (!isBatch) {
-            const outcome = await executeSingleConfigRun({
+            await executeSingleConfigRun({
                 configPath: configPaths[0],
                 options,
                 resolvedOptions
             });
-            if (outcome.shouldFailOnDrift) {
-                console.error(kleur.red('Snapshot eval drift detected in fail_on_drift mode.'));
-                process.exit(2);
-            }
             console.log(kleur.gray('Process exiting.'));
             return;
         }
@@ -160,224 +152,6 @@ program
         process.exit(1);
     }
 });
-program
-    .command('snapshot')
-    .description('Manage evaluation snapshots')
-    .addCommand(new Command('create')
-    .description('Create snapshot from a run (only fully passing runs)')
-    .requiredOption('--run <runId>', 'Run id from runs/<runId>')
-    .option('--name <name>', 'Snapshot name')
-    .option('--runs-dir <path>', 'Directory with run artifacts', 'mcplab/results/evaluation-runs')
-    .option('--snapshots-dir <path>', 'Directory for snapshots', 'mcplab/snapshots')
-    .action((options) => {
-    try {
-        const resultsPath = resolve(options.runsDir, String(options.run), 'results.json');
-        const results = JSON.parse(readFileSync(resultsPath, 'utf8'));
-        const snapshot = buildSnapshotFromRun(results, options.name);
-        const path = saveSnapshot(snapshot, resolve(options.snapshotsDir));
-        console.log(kleur.green(`Snapshot created: ${snapshot.id}`));
-        console.log(kleur.gray(`Path: ${path}`));
-    }
-    catch (err) {
-        console.error(kleur.red(`Error: ${err?.message ?? String(err)}`));
-        process.exit(1);
-    }
-}))
-    .addCommand(new Command('eval-init')
-    .description('Create baseline snapshot from a run and link it to config snapshot_eval policy')
-    .requiredOption('--run <runId>', 'Run id from runs/<runId>')
-    .requiredOption('--config <path>', 'Path to eval.yaml')
-    .option('--name <name>', 'Snapshot name')
-    .option('--runs-dir <path>', 'Directory with run artifacts', 'mcplab/results/evaluation-runs')
-    .option('--snapshots-dir <path>', 'Directory for snapshots', 'mcplab/snapshots')
-    .action((options) => {
-    try {
-        const resultsPath = resolve(options.runsDir, String(options.run), 'results.json');
-        const results = JSON.parse(readFileSync(resultsPath, 'utf8'));
-        const snapshot = buildSnapshotFromRun(results, options.name);
-        saveSnapshot(snapshot, resolve(options.snapshotsDir));
-        const configPath = resolve(String(options.config));
-        const { sourceConfig } = loadConfig(configPath);
-        const nextConfig = {
-            ...sourceConfig,
-            snapshot_eval: {
-                enabled: true,
-                mode: sourceConfig.snapshot_eval?.mode ?? 'warn',
-                baseline_snapshot_id: snapshot.id,
-                baseline_source_run_id: results.metadata.run_id,
-                last_updated_at: new Date().toISOString()
-            }
-        };
-        writeFileSync(configPath, `${stringifyYaml(nextConfig)}\n`, 'utf8');
-        console.log(kleur.green(`Snapshot eval baseline linked: ${snapshot.id}`));
-        console.log(kleur.gray(`Config updated: ${configPath}`));
-    }
-    catch (err) {
-        console.error(kleur.red(`Error: ${err?.message ?? String(err)}`));
-        process.exit(1);
-    }
-}))
-    .addCommand(new Command('eval-policy')
-    .description('Update snapshot_eval policy in a config')
-    .requiredOption('--config <path>', 'Path to eval.yaml')
-    .requiredOption('--enabled <true|false>', 'Whether snapshot eval is enabled')
-    .requiredOption('--mode <warn|fail_on_drift>', 'Snapshot eval mode')
-    .option('--baseline-snapshot <snapshotId>', 'Baseline snapshot id')
-    .option('--baseline-source-run <runId>', 'Source run id used to create baseline')
-    .action((options) => {
-    try {
-        const enabled = String(options.enabled).toLowerCase() === 'true';
-        const mode = String(options.mode);
-        if (mode !== 'warn' && mode !== 'fail_on_drift') {
-            throw new Error('mode must be warn or fail_on_drift');
-        }
-        const configPath = resolve(String(options.config));
-        const { sourceConfig } = loadConfig(configPath);
-        const nextConfig = {
-            ...sourceConfig,
-            snapshot_eval: {
-                enabled,
-                mode,
-                baseline_snapshot_id: options.baselineSnapshot ?? sourceConfig.snapshot_eval?.baseline_snapshot_id,
-                baseline_source_run_id: options.baselineSourceRun ?? sourceConfig.snapshot_eval?.baseline_source_run_id,
-                last_updated_at: new Date().toISOString()
-            }
-        };
-        writeFileSync(configPath, `${stringifyYaml(nextConfig)}\n`, 'utf8');
-        console.log(kleur.green(`Snapshot eval policy updated: ${configPath}`));
-    }
-    catch (err) {
-        console.error(kleur.red(`Error: ${err?.message ?? String(err)}`));
-        process.exit(1);
-    }
-}))
-    .addCommand(new Command('eval-set-scenario')
-    .description('Set or clear a scenario-level snapshot baseline override in a config')
-    .requiredOption('--config <path>', 'Path to eval.yaml')
-    .requiredOption('--scenario <id>', 'Scenario id')
-    .option('--snapshot <snapshotId>', 'Override baseline snapshot id (omit to clear override)')
-    .option('--source-run <runId>', 'Source run id used to create the scenario baseline')
-    .option('--enabled <true|false>', 'Scenario snapshot eval enabled override')
-    .action((options) => {
-    try {
-        const configPath = resolve(String(options.config));
-        const scenarioId = String(options.scenario).trim();
-        if (!scenarioId)
-            throw new Error('scenario is required');
-        const { sourceConfig } = loadConfig(configPath);
-        const scenarios = [...(sourceConfig.scenarios ?? [])];
-        const scenarioIndex = scenarios.findIndex((s) => typeof s === 'object' && s !== null && !('ref' in s) && s.id === scenarioId);
-        if (scenarioIndex < 0) {
-            throw new Error(`Scenario not found in config.scenarios (inline only): ${scenarioId}`);
-        }
-        const current = scenarios[scenarioIndex];
-        if (!current || typeof current !== 'object' || 'ref' in current) {
-            throw new Error(`Scenario not found in config.scenarios (inline only): ${scenarioId}`);
-        }
-        const nextScenarioSnapshotEval = {
-            ...(current.snapshot_eval ?? {}),
-            ...(options.snapshot !== undefined
-                ? { baseline_snapshot_id: String(options.snapshot || '') || undefined }
-                : {}),
-            ...(options.sourceRun !== undefined
-                ? { baseline_source_run_id: String(options.sourceRun || '') || undefined }
-                : {}),
-            ...(options.enabled !== undefined
-                ? { enabled: String(options.enabled).toLowerCase() === 'true' }
-                : {}),
-            last_updated_at: new Date().toISOString()
-        };
-        if (!nextScenarioSnapshotEval.baseline_snapshot_id) {
-            delete nextScenarioSnapshotEval.baseline_snapshot_id;
-        }
-        if (!nextScenarioSnapshotEval.baseline_source_run_id) {
-            delete nextScenarioSnapshotEval.baseline_source_run_id;
-        }
-        if (nextScenarioSnapshotEval.enabled === undefined &&
-            !nextScenarioSnapshotEval.baseline_snapshot_id &&
-            !nextScenarioSnapshotEval.baseline_source_run_id) {
-            scenarios[scenarioIndex] = {
-                ...current,
-                snapshot_eval: undefined
-            };
-        }
-        else {
-            scenarios[scenarioIndex] = {
-                ...current,
-                snapshot_eval: nextScenarioSnapshotEval
-            };
-        }
-        const nextConfig = {
-            ...sourceConfig,
-            scenarios
-        };
-        writeFileSync(configPath, `${stringifyYaml(nextConfig)}\n`, 'utf8');
-        console.log(kleur.green(`Scenario snapshot baseline ${options.snapshot ? 'set' : 'updated'}: ${scenarioId}`));
-        console.log(kleur.gray(`Config updated: ${configPath}`));
-    }
-    catch (err) {
-        console.error(kleur.red(`Error: ${err?.message ?? String(err)}`));
-        process.exit(1);
-    }
-}))
-    .addCommand(new Command('list')
-    .description('List snapshots')
-    .option('--snapshots-dir <path>', 'Directory for snapshots', 'mcplab/snapshots')
-    .action((options) => {
-    try {
-        const snapshots = listSnapshots(resolve(options.snapshotsDir));
-        if (snapshots.length === 0) {
-            console.log('No snapshots found.');
-            return;
-        }
-        for (const snapshot of snapshots) {
-            console.log(`${snapshot.id}  ${snapshot.name}  (run=${snapshot.source_run_id}, created=${snapshot.created_at})`);
-        }
-    }
-    catch (err) {
-        console.error(kleur.red(`Error: ${err?.message ?? String(err)}`));
-        process.exit(1);
-    }
-}))
-    .addCommand(new Command('show')
-    .description('Show snapshot JSON')
-    .requiredOption('--id <snapshotId>', 'Snapshot id')
-    .option('--snapshots-dir <path>', 'Directory for snapshots', 'mcplab/snapshots')
-    .action((options) => {
-    try {
-        const snapshot = loadSnapshot(String(options.id), resolve(options.snapshotsDir));
-        console.log(JSON.stringify(snapshot, null, 2));
-    }
-    catch (err) {
-        console.error(kleur.red(`Error: ${err?.message ?? String(err)}`));
-        process.exit(1);
-    }
-}))
-    .addCommand(new Command('compare')
-    .description('Compare run against snapshot')
-    .requiredOption('--id <snapshotId>', 'Snapshot id')
-    .requiredOption('--run <runId>', 'Run id from runs/<runId>')
-    .option('--format <format>', 'Output format: table|json', 'table')
-    .option('--runs-dir <path>', 'Directory with run artifacts', 'mcplab/results/evaluation-runs')
-    .option('--snapshots-dir <path>', 'Directory for snapshots', 'mcplab/snapshots')
-    .action((options) => {
-    try {
-        const snapshot = loadSnapshot(String(options.id), resolve(options.snapshotsDir));
-        const resultsPath = resolve(options.runsDir, String(options.run), 'results.json');
-        const results = JSON.parse(readFileSync(resultsPath, 'utf8'));
-        const comparison = compareRunToSnapshot(results, snapshot);
-        if (String(options.format) === 'json') {
-            console.log(JSON.stringify(comparison, null, 2));
-        }
-        else {
-            console.log(formatSnapshotComparisonTable(comparison));
-        }
-    }
-    catch (err) {
-        console.error(kleur.red(`Error: ${err?.message ?? String(err)}`));
-        process.exit(1);
-    }
-}));
 program
     .command('results')
     .description('Query evaluation run artifacts for LLM-first workflows')
@@ -680,7 +454,6 @@ program
     .description('Serve MCPLab app frontend and local API bridge')
     .option('--evals-dir <path>', 'Directory for YAML evals', 'mcplab/evals')
     .option('--runs-dir <path>', 'Directory for run artifacts', 'mcplab/results/evaluation-runs')
-    .option('--snapshots-dir <path>', 'Directory for snapshot artifacts', 'mcplab/snapshots')
     .option('--tool-analysis-results-dir <path>', 'Directory for saved tool analysis reports', 'mcplab/results/tool-analysis')
     .option('--libraries-dir <path>', 'Bundle root for reusable servers/agents/test-cases', 'mcplab')
     .option('--port <number>', 'Port to bind', '8787')
@@ -696,7 +469,6 @@ program
                 port: String(options.port),
                 evalsDir: String(options.evalsDir),
                 runsDir: String(options.runsDir),
-                snapshotsDir: String(options.snapshotsDir),
                 toolAnalysisResultsDir: String(options.toolAnalysisResultsDir),
                 librariesDir: String(options.librariesDir)
             })
@@ -705,7 +477,6 @@ program
                 port: String(options.port),
                 evalsDir: String(options.evalsDir),
                 runsDir: String(options.runsDir),
-                snapshotsDir: String(options.snapshotsDir),
                 toolAnalysisResultsDir: String(options.toolAnalysisResultsDir),
                 librariesDir: String(options.librariesDir)
             };
@@ -718,7 +489,6 @@ program
             port,
             evalsDir: resolve(resolvedAppOptions.evalsDir),
             runsDir: resolve(resolvedAppOptions.runsDir),
-            snapshotsDir: resolve(resolvedAppOptions.snapshotsDir),
             toolAnalysisResultsDir: resolve(resolvedAppOptions.toolAnalysisResultsDir),
             librariesDir: resolve(resolvedAppOptions.librariesDir),
             dev: Boolean(options.dev),
@@ -743,13 +513,64 @@ function openBrowserUrl(url) {
         spawn(cmd, [url], { stdio: 'ignore', detached: true }).unref();
     }
 }
+function parseRuntimeServerOverrides(options) {
+    const serverOverrideAll = options.serverOverrideAll
+        ? options.serverOverrideAll
+            .split(',')
+            .map((id) => id.trim())
+            .filter(Boolean)
+        : undefined;
+    if (options.serverOverrideAll !== undefined &&
+        (!serverOverrideAll || serverOverrideAll.length === 0)) {
+        throw new Error('serverOverrideAll must include at least one server id');
+    }
+    const scenarioServerOverrides = {};
+    for (const rawEntry of options.serverOverride) {
+        const entry = String(rawEntry ?? '').trim();
+        const eqIdx = entry.indexOf('=');
+        if (eqIdx < 1) {
+            throw new Error(`Invalid --server-override format '${entry}'. Expected: <scenarioId>=<serverRef[,serverRef...]>`);
+        }
+        const scenarioId = entry.slice(0, eqIdx).trim();
+        const csv = entry.slice(eqIdx + 1);
+        if (!scenarioId) {
+            throw new Error(`Invalid --server-override '${entry}': scenario id cannot be empty`);
+        }
+        const parsedServerIds = csv
+            .split(',')
+            .map((id) => id.trim())
+            .filter(Boolean);
+        if (parsedServerIds.length === 0) {
+            throw new Error(`Invalid --server-override '${entry}': must include at least one server id after '='`);
+        }
+        scenarioServerOverrides[scenarioId] = parsedServerIds;
+    }
+    return {
+        serverOverrideAll,
+        scenarioServerOverrides: Object.keys(scenarioServerOverrides).length > 0 ? scenarioServerOverrides : undefined
+    };
+}
+function filterRuntimeOverridesToSelectedScenarios(selectedConfig, overrides) {
+    if (!overrides.scenarioServerOverrides)
+        return overrides;
+    const selectedIds = new Set(selectedConfig.scenarios.map((scenario) => scenario.id));
+    const filtered = Object.fromEntries(Object.entries(overrides.scenarioServerOverrides).filter(([scenarioId]) => selectedIds.has(scenarioId)));
+    return {
+        ...overrides,
+        scenarioServerOverrides: Object.keys(filtered).length > 0 ? filtered : undefined
+    };
+}
 async function executeSingleConfigRun(params) {
     const { configPath, options, resolvedOptions } = params;
     const loaded = loadConfig(resolve(configPath));
-    const { agents: libraryAgents } = readLibraries(loaded.bundleRoot);
-    loaded.config = { ...loaded.config, agents: { ...libraryAgents, ...loaded.config.agents } };
+    const { agents: libraryAgents, servers: libraryServers } = readLibraries(loaded.bundleRoot);
+    loaded.config = {
+        ...loaded.config,
+        agents: { ...libraryAgents, ...loaded.config.agents },
+        servers: { ...libraryServers, ...loaded.config.servers }
+    };
     loaded.hash = hashConfig(loaded.config);
-    const { config, hash, warnings } = loaded;
+    const { config, warnings } = loaded;
     for (const warning of warnings) {
         console.log(kleur.yellow(`⚠ ${warning}`));
     }
@@ -764,14 +585,18 @@ async function executeSingleConfigRun(params) {
         : requestedAgentsFromCsv.length > 0
             ? requestedAgentsFromCsv
             : undefined;
-    const beforeExpandCount = config.scenarios.length;
-    const effectiveAgents = requestedAgents ?? config.run_defaults?.selected_agents;
-    const expanded = expandConfigForAgents(config, effectiveAgents);
+    const runtimeOverrides = parseRuntimeServerOverrides(options);
+    const selectedBaseConfig = options.scenario ? selectScenarios(config, options.scenario) : config;
+    const selectedOverrides = filterRuntimeOverridesToSelectedScenarios(selectedBaseConfig, runtimeOverrides);
+    const runtimeOverriddenConfig = applyRuntimeServerOverrides(selectedBaseConfig, selectedOverrides);
+    const effectiveConfigHash = hashConfig(runtimeOverriddenConfig);
+    const beforeExpandCount = runtimeOverriddenConfig.scenarios.length;
+    const effectiveAgents = requestedAgents ?? runtimeOverriddenConfig.run_defaults?.selected_agents;
+    const expanded = expandConfigForAgents(runtimeOverriddenConfig, effectiveAgents);
     if (expanded.scenarios.length !== beforeExpandCount || effectiveAgents?.length) {
         const agentCount = effectiveAgents?.length ?? Object.keys(config.agents).length;
         console.log(kleur.cyan(`📊 Testing ${beforeExpandCount} scenarios × ${agentCount} selected agents = ${expanded.scenarios.length} total tests`));
     }
-    const selected = selectScenarios(expanded, options.scenario);
     const runsPerScenario = Number(options.runs);
     if (Number.isNaN(runsPerScenario) || runsPerScenario <= 0) {
         throw new Error('Runs must be a positive number');
@@ -795,9 +620,11 @@ async function executeSingleConfigRun(params) {
         oauthTokens[serverName] = token;
     }
     // Detect OAuth servers missing a token and fail early with a helpful message
-    const oauthServers = Object.entries(config.servers ?? {})
-        .filter(([, v]) => v.auth?.type === 'oauth_authorization_code')
-        .map(([name]) => name);
+    const effectiveServerIds = new Set(expanded.scenarios.flatMap((scenario) => scenario.servers));
+    const oauthServers = Array.from(effectiveServerIds).filter((name) => {
+        const cfg = expanded.servers?.[name];
+        return cfg?.auth?.type === 'oauth_authorization_code';
+    });
     const missingTokenServers = oauthServers.filter((name) => !oauthTokens[name]);
     if (missingTokenServers.length > 0) {
         for (const name of missingTokenServers) {
@@ -814,11 +641,11 @@ async function executeSingleConfigRun(params) {
         }
         throw new Error(`OAuth login required for server(s): ${missingTokenServers.join(', ')}. Provide tokens via --oauth-token.`);
     }
-    const { runDir, results } = await runAll(selected, {
+    const { runDir, results } = await runAll(expanded, {
         runsPerScenario,
         scenarioId: options.scenario,
         runNote,
-        configHash: hash,
+        configHash: effectiveConfigHash,
         gitCommit: getGitCommit(),
         cliVersion: pkgVersion,
         runsDir: String(options.runsDir),
@@ -830,35 +657,6 @@ async function executeSingleConfigRun(params) {
             }
         }
     });
-    let shouldFailOnDrift = false;
-    const useSnapshotEval = Boolean(options.snapshotEval) || Boolean(config.snapshot_eval?.enabled);
-    if (useSnapshotEval) {
-        const policy = config.snapshot_eval;
-        if (!policy?.baseline_snapshot_id) {
-            console.log(kleur.yellow('⚠ Snapshot eval enabled but no baseline snapshot is configured.'));
-        }
-        else {
-            const snapshot = loadSnapshot(String(policy.baseline_snapshot_id), resolve(options.snapshotsDir));
-            const comparison = compareRunToSnapshot(results, snapshot);
-            const enabledScenarioIds = new Set(selected.scenarios
-                .filter((scenario) => scenario.snapshot_eval?.enabled !== false)
-                .map((scenario) => scenario.id));
-            const applied = applySnapshotPolicyToRunResult({
-                results,
-                comparisons: [comparison],
-                policy,
-                enabledScenarioIds
-            });
-            console.log('');
-            console.log(kleur.cyan('📸 Snapshot Eval Policy'));
-            console.log(`${applied.mode} · baseline=${applied.baseline_snapshot_id} · overall=${applied.overall_score} · status=${applied.status}`);
-            if (applied.impacted_scenarios.length > 0) {
-                console.log(kleur.yellow(`Impacted scenarios: ${applied.impacted_scenarios.join(', ')}`));
-            }
-            console.log(formatSnapshotComparisonTable(comparison));
-            shouldFailOnDrift = policy.mode === 'fail_on_drift' && applied.impacted_scenarios.length > 0;
-        }
-    }
     const reportPath = join(runDir, 'report.html');
     const resultsPath = join(runDir, 'results.json');
     const summaryPath = join(runDir, 'summary.md');
@@ -866,19 +664,12 @@ async function executeSingleConfigRun(params) {
     writeFileSync(reportPath, renderReport(results), 'utf8');
     writeFileSync(summaryPath, renderSummaryMarkdown(results), 'utf8');
     console.log(kleur.green(`✅ Run complete. Results: ${runDir}`));
-    if (options.compareSnapshot) {
-        const snapshot = loadSnapshot(String(options.compareSnapshot), resolve(options.snapshotsDir));
-        const comparison = compareRunToSnapshot(results, snapshot);
-        console.log('');
-        console.log(kleur.cyan('📸 Snapshot Comparison'));
-        console.log(formatSnapshotComparisonTable(comparison));
-    }
     const failedRuns = results.scenarios.reduce((sum, scenario) => sum + scenario.runs.filter((run) => !run.pass).length, 0);
     return {
         runDir,
         runId: results.metadata.run_id,
         passed: failedRuns === 0,
-        shouldFailOnDrift
+        shouldFailOnDrift: false
     };
 }
 function getGitCommit() {
@@ -902,9 +693,9 @@ function formatRunProgressEvent(event) {
         case 'run_started':
             return `Run started (${event.totalScenarioRuns} scenario run(s), ${event.runsPerScenario} run(s) each).`;
         case 'mcp_connect_started':
-            return `Connecting MCP servers (${event.serverCount})...`;
+            return `Connecting MCP servers (${event.serverCount}): ${event.serverNames.join(', ')}...`;
         case 'mcp_connect_finished':
-            return `Connected MCP servers (${event.serverCount}).`;
+            return `Connected MCP servers (${event.serverCount}): ${event.serverNames.join(', ')}.`;
         case 'scenario_run_started':
             return `Scenario ${event.scenarioRunIndex}/${event.totalScenarioRuns} started: ${event.scenarioId} [agent=${event.agentName}, run=${event.runIndex + 1}/${event.runsPerScenario}]`;
         case 'scenario_run_finished':