Mikie

com.getmikiePublished by getmikie.comv1.1.166Updated Nov 3, 2026

Focused repository context for Claude Code and Codex, with per-session token usage receipts.

VerifiedSTDIODesktop onlyDeveloper ToolsData & Analytics

Overview

AI-generated overview

Mikie gives coding agents compact, line-anchored repository context packets plus hosted token-usage and quality receipts for each session.

What you can do
  • Build a bounded, line-anchored context packet for a coding-agent prompt from the connected project
  • Rank local repo files and return contiguous implementation and test evidence as a targeted follow-up
  • List Mikie capabilities, reports, data coverage and readiness gaps
  • Read hosted read-only meter, measurement, quality and business summaries
  • Estimate savings from the cost model with optional price and token overrides
When to use it
Useful when you drive Claude Code or Codex on a local repository and want tighter, evidence-anchored context instead of broad file retrieval, and when you want to check token usage, quality comparisons or cost-per-patch style metrics for those sessions.
Requirements
Runs as a local stdio process from the npm package @getmikie/cli on a desktop machine; no environment variables, headers or authentication are declared. Hosted summary tools query a remote read-only service, so network access is implied.
Before you install
The hosted summary tools read from a remote service, so repository-derived or session data may leave your machine; confirm what is sent before use. Context and ranking tools are read-only but expose local repo file contents to the assistant. No credentials are declared, so any authorization the hosted meter needs is undocumented.

Installation

In SourceWeft

  1. Open Mikie in the dashboard and add it to a workspace.
  2. Enable the server for the chats that should use its tools.

Desktop only via STDIO. STDIO servers start a local process, so they need the SourceWeft desktop host.

Tools

8

build_context_packet

Build a compact, line-anchored MikieMCP packet for a coding-agent prompt. Uses the connected project; no cwd argument is needed. Automatically selects a context budget from the task and required evidence. Hard limits: max_files <= 12 and max_chars <= 20000 canonical JSON UTF-8 bytes. Call once for broad discovery. If ordinary reads are allowed, read named gaps directly; do not re-rank known paths. MCP-only: at most five additional targeted Mikie lookups per task; no raw fallback. Reuse complete excerpts. Stop when ready to edit/verify; do not chase unrelated graph rows. Never repeat an unchanged lookup. Stop MCP recovery on an error or no progress; state the gap.

Parameterspromptrequiredmax_charsmax_filescontext_modeallow_noverifyanswerability_scoreexternal_verifier_availableestimated_raw_context_tokensruntime_provisioning_available

rank_relevant_files

Rank local repo files and return bounded contiguous implementation and test evidence. Hard limit: max_files <= 20. Multi-layer traces default to accuracy depth; simple lookups stay compact. Use only as a targeted follow-up. Do not iteratively broaden retrieval within a case; after this result, synthesize or request another distinct missing file within the five-follow-up task budget. For oversized files use prompt path:start-end ranges; combine consecutive ranges without treating an individual range as a complete body. Files include line_count.

Parameterspromptrequiredmax_filescontext_mode

list_mikie_capabilities

List MikieMCP MCP capabilities, reports, data coverage, and readiness gaps.

Parameterslanetask

get_mikie_meter

Query the hosted read-only Mikie Meter through the shared authorization, projection, pagination, and size-bound service. For the standard summary, pass exactly {"view":"summary","authorization_view":"aggregate"}. Do not send an empty or partial argument object.

Parametersviewrequiredlimitcursorend_atperiodfilterstask_idrevisionstart_atreceipt_idexport_formatdimension_names+2 more

get_measurement_summary

Return a hosted compatibility summary projected from the authorized PostgreSQL Mikie Meter at one immutable watermark.

Takes no parameters.

get_quality_summary

Read paired accuracy/quality comparisons for raw/baseline versus MikieMCP.

Takes no parameters.

get_business_summary

Read MikieMCP unit-economics and operational AI-work metrics: cost per accepted patch, cost per resolved issue, cost per merged PR, retry rates, review time, regressions, rollbacks, and useful-versus-garbage agent output.

Takes no parameters.

estimate_mikie_savings

Estimate MikieMCP savings from the cost model with optional price/token overrides.

Parameterscontext_ratiooutput_tokenstool_call_countraw_input_tokensmikie_mcp_output_tokensmikie_mcp_tool_call_countprice_per_1m_input_tokens

Version history

1
  1. v1.1.166LatestOct 30, 2026

Score

55% of the applicable checks met

How complete this listing is, computed from the registry entry, the endpoint and public sources. Not a rating of the server and not a statement that it is safe or working today: reachability is a dated check, and no installs or reviews go into the score.

  • Reached in the last 30 days25 pts
  • Auth declared10 pts

    Declare the credential headers in the registry entry, or keep the endpoint answering the probe so its auth can be classified.

  • Tools listedRequired15 pts
  • DescriptionRequired15 pts
  • Logo10 pts

    Publish an icon in the registry entry.

  • Documentation15 pts

    Declare a GitHub repository with a README, or serve resource_documentation in the OAuth protected-resource metadata.

  • License10 pts

    Declare a license in the repository or the registry entry.