Ennodia isn’t another agent It connects the agents you use
A free MCP server for Codex, Claude Code, and six other agents. When your agent lacks an ability, it calls one that has it and keeps the answer in your chat.
Codex can’t hear audio. In a 30-clip test, it got 10 of 10 speech questions right through Gemini, and 4 of 10 alone. See Fig. 1
Just ask your agent:
Try Ennodia in this conversation: https://try-ennodia.cherninlab.com
Free, open source, and runs on your computer. Set it up yourself
Why use Ennodia
-
Keep the app you like
Ennodia has no window of its own. You work in your agent as before, and it can now reach the others.
-
Your agents work together
Any agent can hand part of a task to any other and get the answer back in your chat.
-
Nothing to pay, nothing to join
Ennodia is free and open source, with no account. It uses the agents and plans you have.
-
It grows with your agents
It is a set of tools, not a framework or a method. Each new model one agent gains becomes an ability for all of them.
-
It stays on your computer
Every handoff leaves a receipt there: who did the work, how long it took, and what came back. Nothing goes to Ennodia.
-
Less to keep up with
You don’t need to follow every new model. Name the task, and your agent hands it to one that can do it.
Next: Ennodia will read its receipts, so your agent stops repeating what failed. Roadmap Questions about setup, data, and cost
What changes when Codex can listen
Codex can’t hear audio. In Fig. 1, each row runs from Codex alone to Codex with Ennodia, which hands the clip to Gemini. Speech rises from 4 to 10, and sound falls from 7 to 5.
- Said it listened
- 0 → 30 of 30
- Median time
- 27 s → 57 s per clip
- Median tokens
- 85,443 → 157,085 per clip
| Task | Codex alone | Codex with Ennodia |
|---|---|---|
| Sound | 7 of 10 | 5 of 10 |
| Music | 6 of 10 | 6 of 10 |
| Speech | 4 of 10 | 10 of 10 |
| All 30 clips | 17 of 30 | 21 of 30 |
Your agent stays in charge
Follow the olive line in Fig. 2. It leaves your agent’s tool call, runs through the agent that does the work, and returns to the same chat. Pick a task to change the route.
Codex GPT-6 Astra
YouGemini, can you listen to this eight-second audio file and tell me what you hear?
Toolennodia_run: Antigravity
Your history stays in Codex
CodexWhat Gemini heard
the file pathwith the model and a time limit
- CodexGPT-6 Astra Draws with GPT Image Fresh session, its own account
- Claude CodeOpus 5.5 Top coding-agent score Fresh session, its own account
- OpenCodeKimi K3 Top long-document score Fresh session, its own account
- AntigravityGemini 3.8 Flash Hears audio files Fresh session, its own account
What Gemini heard, with a receipt
- 1Open Antigravity
- 2Repeat the request
- 3Wait for the answer
- 4Copy it back into Codex
- Through Ennodia
- By hand
- Also reachable
No model leads every task
Fig. 3 names the leader in four kinds of work, as each public leaderboard reports it. A different agent leads each row, so each agent you connect adds a strength. Fig. 4 shows when combining answers helps.
- Coding agents Opus 5.5 in Claude Code 66 index Codex with GPT-6 Astra scores 62. Antigravity with Gemini 3.8 Flash scores 42, at under a fifth of the cost per task. Artificial Analysis Coding Agent Index v1.5
- Long documents Kimi K3 in OpenCode 88.7% Step 5 Preview scores 88.3%, and MiMo-V2.6-Pro 86.3%. Artificial Analysis AA-LCR v1.1
- Image generation GPT Image 2.5 in Codex 1,196 Elo OpenAI holds the top three places. The best Google model, Nano Banana 2, scores 1,123. Artificial Analysis text-to-image arena
- Audio input Gemini in Antigravity Native Current Claude models take text and images only. Gemini audio docsClaude models overview
- In favor Picking among edits from top coding systems scored 66.2% on SWE-bench Verified, above the best single system. CodeMonkeys, 2025
- Against Running the best model many times and combining its answers did 6.6% better on AlpacaEval 2.0 than mixing different models. Self-MoA, 2025
A team helps when each member brings something the others lack, and when the answer is picked well. So Ennodia routes each part to the agent with the ability, and Compare stays optional.
Why not just switch apps?
Both of the usual ways move you out of the agent that knows your project. Read Fig. 5 across a row: where you work, who carries the work, and whose account it uses.