Releases: fullstackhouse/agentloop
Releases · fullstackhouse/agentloop
v0.7.5
v0.7.4
Changes
- Fix: Prevent endless reply loop when Claude Code fails
- Fix: slackUsers filter not working - agent was responding to non-allowlisted users
v0.7.3
Bug Fixes
- slack: Add retry logic and prevent endless error replies when Claude Code fails
What's Changed
When Claude Code execution failed, the error handler posted an error message but never marked the message as processed. On the next poll cycle, the message was re-processed, failed again, and caused an endless loop of error replies.
Now:
- Retries up to 3 times before giving up
- Marks message as processed after max retries to prevent endless loops
- New
maxRetriesconfig option to customize retry count
Full Changelog: v0.7.2...v0.7.3
v0.7.2
Changes
- Fix Slack message formatting — convert markdown to mrkdwn
- Improve Slack thread reply handling
- Update e2e tests
v0.7.1
What's Changed
- Fix Slack message formatting: Convert markdown to Slack's mrkdwn format before posting messages using
slackify-markdownlibrary. This fixes issues where**bold**,*italic*, and[links](url)weren't rendering properly in Slack.
v0.7.0
What's Changed
Features
- Support HTTP-based MCP servers with auth headers
- Reuse previous AI conversation when continuing in the same thread
Other
- Switch e2e tests to #agentloop-test channel
v0.6.0
v0.5.0
What's new
- Agent now has context about the Slack thread it's responding to (full conversation history passed to Claude)
Improvements
- Updated release process: npm publish is now automated via CI (OIDC provenance)
v0.4.2
Add npm publishing with OIDC trusted publishing
v0.4.1
Fix duplicate replies when reactions.get fails