Skip to main content

AgentXchain v2.155.26

v2.155.26 source-backs all nine competitor comparison pages with official documentation links, last-checked dates, and frozen regression tests.

Features​

  • Source-verified comparison pages: every comparison page now includes a public "Source baseline" section with official documentation links and a last-checked date. Readers can verify both competitor and AgentXchain claims against the cited sources. Pages covered: CrewAI, LangGraph, OpenAI Agents SDK, AG2/AutoGen, Devin, MetaGPT, OpenHands, Codegen, and Warp.
  • Factual corrections across five pages: corrected fabricated CLI examples on Codegen (codegen deploy, codegen assign replaced with documented codegen agent create), removed unsupported "smart model routing" claim for Codegen, corrected MetaGPT "greenfield-only" framing (has --project-path incremental dev), split AG2 guardrails/safeguards into distinct features, noted Swarm deprecation in AG2, replaced unsupported exact SWE-Bench 77.6% score for OpenHands with official benchmark-harness scope, and updated Warp from "AI-native terminal" to current Agentic Development Environment / Oz positioning.
  • Regression-guarded source links: each comparison page has a dedicated AT-*-CLAIMS-005 (or equivalent) test that freezes official source links so stale URLs are caught by CI before they reach the public site.
  • Competitive positioning matrix refresh: .planning/COMPETITIVE_POSITIONING_MATRIX.md updated with source links, verification notes, and "What Competitors Do Better" entries for all nine competitors.

Evidence​

  • node --test --test-timeout=60000 cli/test/compare-crewai-claims.test.js cli/test/compare-langgraph-claims.test.js cli/test/compare-openai-agents-sdk-claims.test.js cli/test/compare-autogen-claims.test.js cli/test/compare-devin-claims.test.js cli/test/compare-metagpt-claims.test.js cli/test/compare-openhands-claims.test.js cli/test/compare-codegen-claims.test.js cli/test/compare-warp-claims.test.js cli/test/comparison-pages-content.test.js cli/test/compare-page-architecture.test.js -> 98 tests / 11 suites / 0 failures / 0 skipped

Decisions​

  • DEC-WATCH-DELIVERY-DEDUP-DEFER-001: HTTP-level X-GitHub-Delivery dedup deferred; intake-level signal-hash dedup is sufficient until real operator evidence says otherwise.
  • DEC-OPENAI-AGENTS-SDK-MODEL-BOUNDARY-001: use "configurable model selection" with adapter boundary, not naked "100+ LLMs" claims.
  • DEC-CODEGEN-COMPARE-MODEL-ROUTING-001: use "configurable model selection," not "smart model routing," for Codegen.
  • DEC-CODEGEN-COMPARE-CLI-EXAMPLES-001: only documented CLI commands in Codegen comparison examples.
  • DEC-OPENHANDS-COMPARE-BENCHMARK-SCORE-001: do not freeze unsupported exact SWE-Bench percentages; use official benchmark-harness scope.
  • DEC-DEVIN-COMPARE-WEBHOOK-BOUNDARY-001: use API / webhook-bridge wording for Devin unless official docs expose native webhooks.
  • DEC-WARP-COMPARE-ADE-OZ-BOUNDARY-001: lead with the Warp/Oz split, not just "AI-native terminal."

Tester Re-Run Contract​

Verify the shipped package:

npx --yes -p [email protected] -c "agentxchain --version"
npx --yes -p [email protected] -c "agentxchain dashboard --help"

For comparison page verification, confirm each page at /docs/compare/vs-* includes a "Source baseline" section with official documentation links and a last-checked date.

For downstream closure proof, quote any startup_latency_ms, elapsed_since_spawn_ms, and runtimes.<id>.startup_watchdog_ms evidence relevant to adapter-path reliability while confirming the downstream flow still uses the real runtime path. If a standing phase gate advances before dispatch, quote phase_entered with reconciled_before_dispatch; if continuous mode chains into another run, quote the session_continuation event.

For BUG-55 regression proof, continue using accept-turn followed by checkpoint-turn and quote verification.produced_files, undeclared_verification_outputs, and git status --short.

Clean tree means BUG-55 is fixed for your reproduction.