Repository navigation
Any agent: the aexp CLI and MCP server (story 28.8) - #224
Merged
Merged
Conversation
…8, part 1) AgentExperience.Cli (command aexp): run start/call/end, finalize with the configured verifier, recall of the block and of a failure hint, lessons, export and benefit, and an MCP stdio server with the same operations. Block and hint come from a probe agent wired with UseAgentExperience, so they are the library's own; capture and finalization drive Core directly. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01GPXz56QPQAzDSyJTxbQTTy
…over MCP (story 28.8) Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01GPXz56QPQAzDSyJTxbQTTy
…P (story 28.8) Eight packages: verify-packages checks a tool package (its command, tools/ layout, bundled libraries at this release), release.yml counts eight, the public API baseline gains the tool's (empty) surface. Tool references are not library floors. Guide page, README section, CHANGELOG, CONTRIBUTING. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01GPXz56QPQAzDSyJTxbQTTy
…tory 28.8) Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01GPXz56QPQAzDSyJTxbQTTy
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01GPXz56QPQAzDSyJTxbQTTy
… and errors (story 28.8) aexp trust approves a config's exact bytes in the user's own config dir; the verifier runs (and a reported exit code counts) only for a trusted config. The verifier's program is resolved by aexp (PATH only, or inside the project). Discovery stops at the repository root or home; a config another user owns is refused. finalize checks the run before spawning anything; experience_finish keeps an ended run for a retry. Journals are written atomically under a lock file; task text is sanitized; runs prune. Written files stay inside the project; MCP errors are fixed messages. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01GPXz56QPQAzDSyJTxbQTTy
…ournals (story 28.8) Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01GPXz56QPQAzDSyJTxbQTTy
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Story 28.8 of Epic 28. A new dotnet tool,
AgentExperience.Cli(aexp), lets any agent use the library: a CLI and an MCP stdio server over the SQLite store (or PostgreSQL).run start|call|end,finalize(the configured verifier, in a later process if need be),recall(the block, or the failure hint),lessons,export,benefit,runs prune,trust.aexp mcp):experience_recall,experience_failure_hint,experience_record_call,experience_finish,experience_lessons; bounded output; fixed error messages to the model.ChatClientAgentwithUseAgentExperienceover a scripted client, no model call), so writer, eligibility, recall bounds and holdout are the adapter's; capture and finalization call Core directly with the library's sanitizer..aexp/config.json(discovery stops at the repository root). A config may drive the verifier or accept an exit code only after a human runsaexp trust, which records its hash outside the project; an edited config is refused until re-approved. The verifier program is resolved from PATH or inside the project, never the current directory; finalize validates the run before spawning anything; journals are written atomically; export paths are confined to the project.InternalsVisibleTofrom the MAF package to the tool (it bundles that exact assembly);ModelContextProtocol.Core2.3.0. No public API change. Guide: docs/guide/cli-and-mcp.md.Tests: Cli.Tests 44 (in-process and the packed tool over real stdio), MAF 1458, Release 199, every Docker-free project; pack and verify-packages (8 packages).
🤖 Generated with Claude Code
https://claude.ai/code/session_01GPXz56QPQAzDSyJTxbQTTy