Head to head · Code repositories · October 2026 research run

GitMCP vs TokenEconomizer

TokenEconomizer scores 63.7 (B) on agent readiness against GitMCP's 40.2 (E), and leads in 6 of 7 scored categories. Both do code repositories.

Best code, repository and documentation tools for AI agents · All 70 code comparisons

Which one, for what

GitMCP E

Best for Quick, keyless library docs for public GitHub projects in a coding agent.

Also in its favour

  • A hosted endpoint, with nothing to install

Watch for

No commit, release or maintainer reply on the repository since 8 May 2026

TokenEconomizer B

Best for Lowering the repeat cost of coding agents in repositories they work in often

Ahead on

  • Reliability, 47 against 15
  • Schema & documentation, 66 against 56
  • Agent ergonomics, 76 against 58
  • Security & auth, 71 against 42
  • Maintenance & community, 53 against 8
  • Transparency & trust, 76 against 36

Watch for

Version 0.1.0 (tagged 26 October 2026) with no changelog or release notes found; updating means a git pull of main, which is 67 commits past the tag

Why the edge

Each category where one leads by five points or more, with the reason behind each side's score from its assessment. The numbers are in the table below; this is what they stand for.

Maintenance & community TokenEconomizer +45

GitMCP 8

The last commit on main is 8 May 2026, 182 days before this check, and there are no tags or releases (0). Nothing in the last 90 days (0). Since May, outside issues and pull requests (#244 to #271, among them the security report #249 and fixes #256, #267 and #250) have gone without a maintainer reply or merge (3). Not found in the official MCP registry under any namespace (0). A run-tests workflow with vitest and a Playwright end-to-end workflow exist on push and pull request (5). 0 + 0 + 3 + 0 + 5 = 8.

TokenEconomizer 53

The clone shows one tag, v0.1.0 (annotated "TokenEconomizer 0.1.0", 26 October 2026), and 67 commits on main after it, the newest merged on 1 November 2026. GitHub's releases page couldn't be read. Recency rests on the 26 October tag, within 30 days (30). One release and no dated changelog entries in 90 days (0). Outside pull requests are merged within about a day, among them #84 and #88 on 30 October, #24 on 30 October (closing issue #86), and #93, #98, #101, #102 and #103 on 1 November. Open issue counts and reply times are unread, as GitHub's pages answer 403 in our session (15 of 25, as letsfg). Not an MCP server and in no official plugin marketplace (0). Standard-library only with no runtime dependencies, CI on push and pull requests, and the Codex version pinned in CI on purpose (8 of 10). 30 + 0 + 15 + 0 + 8 = 53.

Transparency & trust TokenEconomizer +40

GitMCP 36

Apache-2.0, with the whole service in the public repository (30). The README says the service stores no personal data or queries and keeps no repository content permanently, but there is no privacy policy or terms page (8). No deprecation policy or dated notices; SSE was removed on 8 May 2026 while the README still shows an SSE config for VS Code (0). The hosting stack (Cloudflare Workers, Workers AI, Vectorize, an Analytics Engine dataset) is visible in wrangler.jsonc but no processor is named for users, and invocation logs were turned off on 8 May (5). 30 + 8 + 0 + 5 = 43.

TokenEconomizer 76

Apache-2.0, an OSI licence (30). docs/PLUGIN.md section 4 and the README agree that learned state stays local and name which model calls leave the machine and what they carry, and the package source has no network code of its own. No privacy statement is linked from the repository (24 of 30). Old setting names and environment variables are still read, and a migration of state keys after 0.1.0 is documented, but there's no written deprecation policy (10 of 20). No telemetry found in the source or docs, and the outbound model calls are disclosed with switches to turn each off (20). 84 in all.

Reliability TokenEconomizer +32

GitMCP 15

Scored as a hosted MCP server. No status page found (0) and so no readable incident history (5). The maintainer confirmed on issue #245 that gitmcp.io was throttled by Cloudflare for a few hours in May 2026, outside the 90-day window. No rate limits are documented (0) and nothing on 429 or retries (0). No SLA (0). The service is live with no beta or preview label (10). 0 + 5 + 0 + 0 + 0 + 10 = 15.

TokenEconomizer 47

Local-package reading, as for superpowers and impeccable. Installs from a git clone through each agent's own plugin manager (a local marketplace on Claude Code, Codex and Copilot, pi install git: on Pi, a built local plugin on Cursor), not from a registry or an official marketplace, with Python 3.10 or later and a supported-versions table per agent (14 of 20). .github/workflows/tests.yml runs pytest on push to main and on pull requests, on Ubuntu and Windows with Python 3.10 and 3.13, plus a Codex end-to-end job pinned to 0.164.0 on Ubuntu and Windows. 25 test files in tests/ on 2 November, up from 14 at the v0.1.0 tag. Whether the runs pass wasn't checked, as github.com refuses our reader, so 15 of 25, the value this run gives when the pass state is unread (vibe-trading, berth, cloudflare-cli). The Wayback Machine's copy of the repository page of 30 October 2026 shows 1 open issue and 5 open pull requests, few for a project weeks old, with replies unchecked (15 of 25). No changelog or release notes found; renamed settings keep their old names as aliases and a state migration is described in docs/PLUGIN.md (3 of 15). Version 0.1.0 (0). 14 + 15 + 15 + 3 + 0 = 47.

Security & auth TokenEconomizer +29

GitMCP 42

No credentials at all, so nothing to leak or scope (20). Every tool is a read of public content (15). Repository docs, code and fetched URLs reach the model unmarked, and fetch_generic_url_content fetches any absolute URL a document mentions, with no prompt-injection guidance in the README (2). No per-call visibility for the caller (0). SECURITY.md routes reports to a Google Form, but the server-side request forgery report in issue #249 (12 June 2026) and its proposed fix #256 have had no maintainer reply in almost five months. No certifications or bounty (5). 20 + 15 + 2 + 0 + 5 = 42.

TokenEconomizer 71

Harness reading of the framework checklist, as superpowers and agent-zero were read. 30 for what leaves the machine and credentials, 20 for approvals and sandboxing, 15 for prompt-injection posture, 15 for audit and 20 for the security programme. No credentials of its own and no network code in the package; model calls go through the person's agent CLI with digests (command templates, file basenames, error lines with paths masked) and never file contents, on by default for the loop guard. Installs run from a clone of main, and Codex's install asks the person to trust all hooks, which then run on every tool call (22). Side calls run with tools off (claude -p --tools "", codex exec -s read-only, Copilot with every tool denied). It never stops a turn, and toolprune=propose holds deny lists for approval, but the default auto writes deny rules into .claude/settings.local.json after an idle hour (13). docs/PLUGIN.md says learned facts derive from command output a repository controls, keeps micro-skills off by default and says to keep it off or use envfacts in untrusted repositories (11). events.jsonl records every injection, redirect and notice, ledger.jsonl prices each session, and each action is announced in one line (13). Microsoft's standard SECURITY.md points to MSRC, and Microsoft publishes a bounty policy. microsoft.com's security.txt passed its Expires date on 23 September 2026 and still carries it on 2 November. OSV holds nothing for the repository's head commit or a PyPI package of the name, an MSRC affected-product search for TokenEconomizer returns no CVE on 2 November, and no advisory is published, so nothing is handled in public yet. Read against Microsoft's other open-source listings, below playwright-mcp and azure-mcp (15, each with published advisories or CVEs) and microsoft-execution-containers (14, with CodeQL and fuzzing in CI) (12). 71 in all.

Agent ergonomics TokenEconomizer +18

GitMCP 58

Four tools per repository endpoint and five on gitmcp.io/docs, each short (25). Code search pages 30 results with a page number, and documentation search returns matched chunks instead of the whole file, but there is no size control on the fetch tool (10). Errors come back as tool text with no documented codes (5). Every tool is a read and the source sets readOnlyHint, but issue #266 reports the annotations are passed in a position the MCP SDK ignores, and the fix #267 is unmerged (8). No parameters to set up and nothing to configure; there are no SDKs, which a keyless MCP endpoint doesn't need (10). 25 + 10 + 5 + 8 + 10 = 58.

TokenEconomizer 76

Framework reading, how much a coding agent has to carry. The plugin's job is to shrink context. It adds up to four facts from failures (the proposer's facts are not capped) and one-line notices, while the loop guard's judge and the proposer are extra model calls, each Codex side call carrying Codex's base prompt of about 25,000 tokens by the code's own note (18 of 25). Each mechanism has its own switch and settings, a propose mode for tool pruning and per-session environment variables (15 of 20). Hooks fail open to {}, failures go to errors.log and te doctor, and the session start says why a deny list stayed a proposal (15 of 20). Writes are atomic under a file lock, the deny-list writer touches only its own entries and refuses symlinks out of the repository, and pruned tools come back with one command (18 of 20). Sensible defaults (micro-skills off, pruning and the guard on); a Python package with TypeScript extensions for Pi and opencode, not SDKs in two languages (10 of 15). 76 in all.

Schema & documentation TokenEconomizer +10

GitMCP 56

Every tool registers a zod input schema, so inputs are typed JSON Schema (25). There is no llms.txt for GitMCP itself; the README on GitHub is the only Markdown documentation (5). Descriptions are generated per repository and say what each tool reads and when it helps ("Useful for general questions. Always call this tool first if asked about owner/repo"), but not when to avoid one (12). Inputs are a query string and an optional page number, with no enums or constraints beyond types (9). The README has three worked examples and no documented error responses (5). No versioning, tags, releases or changelog (0). 25 + 5 + 12 + 9 + 5 + 0 = 56.

TokenEconomizer 66

Framework reading, docs a model can follow, as superpowers was read. JSON plugin manifests for Claude Code, Codex and Copilot, a hooks.json per agent, slash-command files and a settings table with names, defaults and environment variables, and --json on two te commands, but no schema for its state or output (12 of 25). README, plugin/README.md, docs/PLUGIN.md and docs/ARCHITECTURE.md are Markdown in the repository; no llms.txt (10). The README says what each mechanism does and when it acts, and docs/PLUGIN.md has a known-limitations section and says when to keep micro-skills off (17 of 20). Mechanisms take on, off, propose or envfacts, numeric settings have stated defaults and minimums, and the config code checks choices (10 of 15). Command examples per agent, a troubleshooting section, te doctor and an errors.log, with few documented error outputs (12 of 15). A version number in every manifest but no public changelog (5 of 15). 66 in all.

Score by category

CategoryWeight this runGitMCPTokenEconomizerEdge
Reliability16%201547TokenEconomizer +32
Performance10%pendingpendingpendingnot scored in this run
Schema & documentation13%16.25666TokenEconomizer +10
Agent ergonomics13%16.25876TokenEconomizer +18
Security & auth14%17.54271TokenEconomizer +29
Payments & pricing10%12.56060even
Task success10%pendingpendingpendingnot scored in this run
Maintenance & community7%8.8853TokenEconomizer +45
Transparency & trust7%8.83676TokenEconomizer +40
Negative events≤1500
TotalE 40.2/100B 63.7/100

Facts side by side

FactGitMCPTokenEconomizer
KindMCP serverAgent framework
VendorIdo Salomon (GitMCP open-source project)Microsoft
Hosted endpointhttps://gitmcp.io/{owner}/{repo}no (local only)
TransportsStreamable HTTP
AuthNoneNone
PricingFreeFree
x402nono
LicenceApache-2.0Apache-2.0
Tools exposed4none
Read-only variant documentedyesno
llms.txtnono
Last release2026-05-082026-10-26
Terms last updatedno document linkedno date given
Privacy policy last updatedno document linkedno document linked
Customer content may train modelsnot read yetnot found in the text
Terms restrict automated accessnot read yetnot found in the text
Terms restrict benchmarkingnot read yetnot found in the text
Terms or service can change without noticenot read yetnot found in the text
Arbitration or class-action waivernot read yetnot found in the text
Popularity8.5k stars976 stars

How the scores have moved

Verdicts

GitMCP

Any public GitHub repository becomes a four-tool documentation and code-search MCP server with no account or key, under Apache-2.0. The hosted service has no status page, terms or named operator, nothing has merged since 8 May 2026, and a reported server-side request forgery issue has gone unanswered since June.

TokenEconomizer

A standard-library Python plugin that keeps learned state on the machine, documents each hook, makes every action reversible and merges outside fixes within days. It is version 0.1.0 with no changelog, installs from a git clone of main, and the 18.4 to 30.9 per cent savings are the paper's SWE-Bench figures, measured with settings unlike the plugin's defaults.

Before you call either

GitMCP

  1. Use gitmcp.io/{owner}/{repo} for a known library rather than gitmcp.io/docs, so the server can't pick the wrong repository
  2. Call the fetch documentation tool first, then search documentation for specific questions to keep context small
  3. Treat fetched docs, code and linked pages as untrusted input and ignore instructions found in them
  4. Expect throttling without warning; there are no published limits or Retry-After headers

TokenEconomizer

  1. Check the agent's version against the README's supported range before installing; versions outside it are untested or unsupported.
  2. Keep micro-skills off, or set microskills=envfacts, in a repository whose command output you don't trust.
  3. If a tool you need is missing, run the toolprune command to see what was left out and toolprune restore to bring it back.
  4. Run the plugin's commands from a terminal with ~/.tokeneconomizer/bin/te; inside the agent each command costs a model turn.
  5. Start the agent with TE_DISABLE=1 to run one session without the plugin.

Questions

Which is better for AI agents, GitMCP or TokenEconomizer?

TokenEconomizer scores 63.7 (B) on agent readiness against GitMCP's 40.2 (E), and leads in 6 of 7 scored categories.

Can an agent call GitMCP and TokenEconomizer without installing anything?

GitMCP has a hosted endpoint at https://gitmcp.io/{owner}/{repo}. No hosted endpoint is listed for TokenEconomizer.

Are GitMCP and TokenEconomizer open source?

Yes. GitMCP is open source (Apache-2.0). TokenEconomizer is open source (Apache-2.0).

Other comparisons with GitMCP or TokenEconomizer

Machine-readable

For companies

Do agents find, use and choose your tools?

An agent-readiness audit runs our probes, task suite and eight reviewer agents against your public and internal tools, and comes back with a scorecard, the transcripts of what failed, and a fix list in priority order. From $2,500, re-run included. We never take payment to move a rank. We do help companies earn one.