Integrate
Genome over MCP
The genome MCP endpoint, its six tools and a client configuration. It reads and compiles a genome. It cannot import one.
Status
- Not available yet Import a pack through MCP. The genome MCP endpoint has six tools and none imports a file.
- Needs an IntelliNix workspace with the genome flags on Read, compile or export the genome of an agent already in your workspace. Needs the genome_public_api and agent_genome flags and an API key with the genome scopes.
No tool imports a file
The endpoint has the six tools below. None of them imports an AGF file. To install a pack, import it in IntelliNix (see the IntelliNix guide). After that, an MCP client can read or compile the genome of an agent in the workspace.
Prerequisites
- An IntelliNix workspace with
genome_public_apiandagent_genomeon. - An API key with explicit genome scopes (
genome:read,genome:write,genome:erase). Older keys without a scopes field are refused. - Your API address as
[YOUR-INTELLINIX-API]and the key as[YOUR-API-KEY]. - The slug of an agent in your workspace, as
[YOUR-AGENT-SLUG].
Client configuration
The endpoint is POST /mcp/genome. It speaks JSON-RPC 2.0 and answers each request with one JSON
response (no event stream). The key goes in the X-API-Key header, or as a bearer token. The server reports the name
intellinix-genome.
{
"mcpServers": {
"intellinix-genome": {
"url": "[YOUR-INTELLINIX-API]/mcp/genome",
"headers": {
"X-API-Key": "[YOUR-API-KEY]"
}
}
}
}
The URL and the header are the platform's. The outer key names (mcpServers, url,
headers) are a common shape and differ between MCP clients, so check your client's own documentation. It was
not tested with a specific client.
The exact calls
List the tools, then read the genome of an agent.
curl -sS -X POST '[YOUR-INTELLINIX-API]/mcp/genome' \
-H 'X-API-Key: [YOUR-API-KEY]' \
-H 'Content-Type: application/json' \
-d '{"jsonrpc":"2.0","id":1,"method":"tools/list"}' curl -sS -X POST '[YOUR-INTELLINIX-API]/mcp/genome' \
-H 'X-API-Key: [YOUR-API-KEY]' \
-H 'Content-Type: application/json' \
-d '{"jsonrpc":"2.0","id":2,"method":"tools/call","params":{"name":"genome_get","arguments":{"agent":"[YOUR-AGENT-SLUG]"}}}' The tools
| Tool | Key scope | Arguments | What it does |
|---|---|---|---|
genome_get | genome:read | agent | The genome of an agent, each gene with its layer of origin. |
genome_compile | genome:read | agent, task; optional dialect, context, subject_ref, budget_chars | Compile the genome for a task. Returns a compile record id. |
genome_finalize | genome:read | compile_id, output | Strip the genome report from a model output. The output is not stored. |
genome_record_outcome | genome:write | compile_id, verdict (accepted or corrected); optional feedback, edited | Record the outcome of a compile, once. |
genome_export | genome:read | agent | Export an agent’s genome as an AGF document. Administrators only. |
genome_erase | genome:erase | subject_ref | Erase every event and compile record derived from one subject. |
Expected results
tools/list: a result with the six tools, each with its name, description and input schema.genome_get: a result whose content is the agent's composed genome as JSON text.
Troubleshooting
| Error | What it means |
|---|---|
-32001 | Not authenticated. |
-32003 | The key lacks the scope, is an older key without scopes, or the caller may not export. |
-32004 | The agent or compile record was not found, or belongs to another workspace. |
-32029 | Rate limited. The HTTP status is 429 with a Retry-After header. |
-32602 | A tool name or an argument is wrong. The message names the field, not the value. |
HTTP 404 | The genome_public_api or agent_genome flag is off for the key’s workspace. This is a real HTTP 404, for every method. |