2.0.1: Renamed tools, deeper analysis, and faster leaner responses

What's Changed

Upgrading from 1.x? Every tool is renamed, and --allowed-orgs is gone. See Migrating from 1.x.

Added

• Report where the time went by debug log category, with the level each category was captured at, and governor limits by namespace, in apexlog_get_summary (#62, #86, #108, #191)
• Report the debug level each debug log category was captured at, in apexlog_list_slow_operations and apexlog_list_limit_risks, so you can see what the log did not contain (#102, #138)
• Report what ended a failed transaction, with its exception message, and which debug log categories the platform truncated, in apexlog_get_summary (#97, #100)
• Group operations in apexlog_list_slow_operations by namespace, caller namespace or debug log category, so you can see which package or which kind of work the time went to (#101, #126, #127, #131, #138)
• Rank operations by the heap they retain in apexlog_list_slow_operations, not only by time (#99, #127, #138)
• Return the query optimiser's plan for each query apexlog_list_slow_operations ranked (#120)
• Report progress while apexlog_execute_anonymous connects, sets the trace flag, runs and writes, and report when the org logged at levels other than the ones asked for (#65)
• Add --no-apex-execution, which stops apexlog_execute_anonymous running Apex while the log analysis tools keep working (#52)
• Support the 2026-07-28 protocol revision - clients on the 2025 revisions keep working (#103)

Changed

• Breaking: rename every tool with an apexlog_ prefix - analyze_apex_log_performance is now apexlog_list_slow_operations. Update your permission lists and prompts; the full mapping is in Migrating from 1.x (#107)
• Breaking: refuse to run Apex from apexlog_execute_anonymous for production orgs or orgs whose type cannot be determined, unless you confirm. Clients name the org and show the Apex. Allow with --allow-production-orgs. Sandbox, scratch, trial and Developer orgs run unprompted (#52, #93)
• Rank every timed operation by self time in one apexlog_list_slow_operations list - queries, DML, callouts and the rest, not only methods, with repeats of the same name folded into one row, where analyze_apex_log_performance listed the ten slowest single method calls (#86, #97, #108, #126)
• Report all thirteen governor limits in apexlog_get_summary, zeros included where 1.x reported only those above zero, and state each limit's unit (#62, #86, #108, #167)
• Report the limits nearest their ceiling in apexlog_list_limit_risks as one table, beside the threshold that selected it, in place of the four overlapping sections of find_performance_bottlenecks (#108)
• Cut tool responses with no fact lost - apexlog_list_limit_risks by 58%, and apexlog_list_slow_operations ~2.2× smaller, returning fewer rows than asked for where the page would be too large, and stating how many rows matched (#63, #97, #108, #109, #120, #138)
• Reduce the tool definitions token cost on every request by 19%, ~1,529 to ~1,232. Clients can also cache them for an hour (#87, #94, #99, #101, #103, #126, #127, #138, #188, #189)
• Start the server in 55 ms, down from 290 ms, and answer a second question about the same log without parsing it again (#88, #165)

Fixed

• Fix incorrectly reported timings for callouts in apexlog_list_slow_operations, which were counted in the calling method's self time (#97, #138)
• Fix understated governor limit usage in apexlog_get_summary and apexlog_list_limit_risks, which reported the usage the transaction ended on rather than its peak (#97)
• Fix the understated log size in apexlog_get_summary for a log holding non-ASCII characters
• Return the debug log of the run apexlog_execute_anonymous made, where the newest log for the user could be another process's (#65)
• Return an absolute log path from apexlog_execute_anonymous, and warn when outputDir resolves outside every folder the client opened (#109)
• Name the real reason a log file cannot be opened - a permission error used to read as "Log file not found" (#109)
• Mark apexlog_execute_anonymous destructive, so clients stop running it unprompted (#52)

Removed

• Breaking: remove --allowed-orgs and its special tokens - the flag is accepted, ignored and warns, and the org's type decides instead (#52)
• Breaking: drop Node.js 20 (end of life April 2026) - Node.js 22 is the minimum
• Remove the canned recommendations of analyze_apex_log_performance - an agent advises better from the numbers (#86)

New Contributors

• @lukecotter made their first contribution in #84

Full Changelog: 1.0.0...2.0.1

2.0.0-beta.2

What's Changed

Fixed

• Fix the understated size of a log holding non-ASCII characters, and the heap a row retains where the code freed memory, on apex-log-parser 0.1.1

Full Changelog: 2.0.0-beta.1...2.0.0-beta.2

2.0.0-beta.1

What's Changed

Upgrading from 1.x? Every tool is renamed, and --allowed-orgs is gone. See Migrating from 1.x.

Added

• apexlog_list_slow_operations folds repeat calls into one row - by name, by callerNamespace, or by debug log category - each row totalling what the transaction saves if the group never runs, so the rows do not add up ([#101], [#126], [#127], [#131], [#138])
• apexlog_list_slow_operations sorts by an operation's net heap allocation, not only time (sortBy: "heapSelfNetBytes") ([#99], [#127], [#138])
• apexlog_list_slow_operations returns the query optimiser's plan for each query it ranked ([#120])
• apexlog_list_slow_operations returns matchedCount, so you can tell whether the page hid rows ([#63])
• apexlog_list_slow_operations and apexlog_list_limit_risks report the level each debug log category was logged at ([#102], [#138])
• apexlog_execute_anonymous reports progress, and says when the org logged at levels other than the ones asked for ([#65])
• --no-apex-execution stops apexlog_execute_anonymous running Apex, while the log analysis tools keep working ([#52])

Changed

• Breaking: rename every tool with an apexlog_ prefix: analyze_apex_log_performance is now apexlog_list_slow_operations. Full mapping in Migrating from 1.x ([#107])
• Breaking: apexlog_execute_anonymous refuses a production org, or one it cannot classify, unless you confirm the run or set --allow-production-orgs ([#52], [#93])
• apexlog_list_slow_operations ranks every timed operation by self time - a callout or a query as well as a method - in place of the five lists and the prose advice of analyze_apex_log_performance ([#86], [#97], [#108], [#126])
• apexlog_get_summary reports all thirteen governor limits, zeros included, plus time by category and limits by namespace, so a managed package spending your CPU time is visible ([#62], [#86], [#108])
• apexlog_get_summary and apexlog_list_limit_risks report each governor limit's peak, not the figure the transaction ended on ([#97])
• apexlog_list_limit_risks returns one table of the limits nearest their ceiling, covering thirteen where find_performance_bottlenecks covered six ([#108])
• apexlog_get_summary says whether the log was truncated, by what, and how much it skipped ([#100])
• apexlog_list_slow_operations renames its parameters and row fields - topMethods is now limit, minDuration is minSelfMs, and kind is the platform's debug log category and event type, database,SOQL_EXECUTE_BEGIN - and takes a list in every filter ([#86], [#97], [#108], [#138])
• apexlog_execute_anonymous renames success to succeeded, and reports a duration that agrees with apexlog_get_summary ([#65], [#109])
• apexlog_list_slow_operations caps a response by size, not by row count, so one huge log cannot flood the reply: the biggest of 124 real logs returns 15,511 tokens, down from 35,520 ([#108])
• Cut tool responses with no fact lost - apexlog_list_limit_risks by 54%, apexlog_execute_anonymous by 30%. apexlog_get_summary costs 24% more, for the two tables it gained ([#62], [#86], [#97], [#108], [#109], [#120], [#138])
• Cut the cost of having the server connected by 9%, and let a client cache the tool definitions for an hour, though apexlog_list_slow_operations costs 145% more for what it now selects and ranks ([#87], [#94], [#99], [#101], [#103], [#126], [#127], [#138])
• The server starts in 55 ms, down from 290 ms, and reuses the log it parsed, so a second question about the same file skips the parse ([#88], [#165])

Fixed

• apexlog_execute_anonymous returns the log for the Apex you ran, not the newest log for the user ([#65])
• apexlog_execute_anonymous returns an absolute log path, and warns when outputDir lands outside the folders the client opened ([#109])
• The log analysis tools name the real reason a log file cannot be opened. A permission error used to read as "Log file not found" ([#109])
• apexlog_execute_anonymous is marked destructive, so clients stop running it unprompted ([#52])

Removed

• Breaking: remove --allowed-orgs and its special tokens. The flag is accepted, ignored and warns, and the org's type decides instead ([#52])
• Breaking: drop Node.js 20 (end of life April 2026). Minimum node version is 22

New Contributors

• @lukecotter made their first contribution in #84

Full Changelog: 1.0.0...2.0.0-beta.0

1.0.0: Performance analysis, log summaries, bottleneck detection, and anonymous Apex execution with org allowlists and configurable debug levels...

What's Changed

Added

• Performance Analysis (analyze_apex_log_performance) - Feed in a debug log and instantly see which methods are the slowest. See execution times, SOQL/DML counts, and SOSL queries. All durations in milliseconds. Includes log size, debug levels, and thrown exception count.
• Log Summaries (get_apex_log_summary) - Get a debug log summary. Total execution time, method count, governor limit usage (all limits with usage > 0), and log issues as structured {type, summary} objects.
• Bottleneck Detection (find_performance_bottlenecks) - Detects CPU, database and method performance issues by type so you know exactly what to focus on. Empty sections are omitted for cleaner responses.
• Anonymous Apex Execution (execute_anonymous) - Run Apex against any Salesforce org. The debug log is saved to a local file (default: .apex-log-mcp/ in the project root) and a summary with the file path is returned. Use the file path with the analysis tools for deeper investigation. Specify a target org by alias or username, or use the project default.
  • Org allowlist (--allowed-orgs) — Disabled by default, must be explicitly enabled. Supports special tokens: ALLOW_ALL_ORGS (permit any org), DEFAULT_TARGET_ORG and DEFAULT_TARGET_DEV_HUB (resolve from Salesforce CLI config). Aliases in the allowlist are resolved to usernames for matching.
  • Debug levels — Configurable via the debugLevel parameter. Set all categories at once (e.g. "FINEST"), reset to defaults, or override specific categories like apexCode, database, and nba.
  • Output directory — Configurable via the outputDir parameter. Defaults to .apex-log-mcp/ in the project root.

New Contributors

• @ntaylorcertinia made their first contribution in #1
• @pballcertinia made their first contribution in #13
• @mfakhoricertinia made their first contribution in #12
• @lcottercertinia made their first contribution in #22

Full Changelog: https://github.com/certinia/debug-log-analyzer-mcp/commits/1.0.0