Agent manifests
An agent manifest is a structured configuration file that contains metadata about the AI agent, including its capabilities and configuration. It helps the AI system understand what the agent can do and how it should work.
What the file contains
The following are examples of what a manifest contains:
- Agent name and description to identify the agent and its purpose
- LLM model details
- Inputs, outputs, and supported actions
When to use a manifest
The manifest is useful in the following situations:
- Get an overview of the agent's capabilities
- Integrate agents with APIs and external systems such as CRMs
- Share agent specifications with other developers
Example manifest
json
1
2{
3 "functions": [
4 {
5 "name": "getLLMs",
6 "description": "Gets response",
7 "method": "GET",
8 "uri": "https://api.example.com/health",
9 "inSchema": {
10 "properties": {
11 "LLM": {
12 "examples": [
13 "llama-3.2-3b-instruct-gpu",
14 "llama-3.2-1b-instruct-cpu",
15 "Gemini1.5-flash",
16 "GPT4o-mini"
17 ],
18 "title": "LLM",
19 "type": "string"
20 },
21 "Prompt Compression": {
22 "examples": [
23 true
24 ],
25 "title": "Prompt Compression",
26 "type": "object"
27 }
28 },
29 "required": [
30 "LLM"
31 ],
32 "type": "object"
33 },
34 "outSchema": {
35 "pattern": "^(.+)$",
36 "properties": {
37 "LLM": {
38 "default": "",
39 "examples": [
40 "llama-3.2-3b-instruct-gpu",
41 "llama-3.2-1b-instruct-cpu",
42 "Gemini1.5-flash",
43 "GPT4o-mini"
44 ],
45 "pattern": "^(.+)$",
46 "title": "response test",
47 "type": "string"
48 },
49 "Prompt Compression": {
50 "default": "false",
51 "examples": [
52 true,
53 false
54 ],
55 "title": "response test",
56 "type": "string"
57 }
58 },
59 "required": [
60 "response"
61 ],
62 "title": "The Items Schema",
63 "type": "object"
64 }
65 }
66 ]
67}
To view the manifest of your AI agent, refer to View agent manifest.