agami-core
The governed trust layer between an AI agent and your database. The importable core behind the
agami Claude Code skill and MCP server — the semantic model, the shared MCP
TOOLS harness, and the local query executor.
What it does
Point an AI agent at a database and it answers by guessing — at the join, at what "revenue" means, at which rows it's allowed to read. agami-core turns your schema into a governed semantic model where every join is FK-derived or human-approved and every metric is signed off before the runtime trusts it. Every answer then ships a receipt — the exact SQL it ran and the model version it pinned — so a silent join mistake never reaches you as a confident wrong number.
It runs locally: credentials, schema, and query results never leave your machine.
Most people don't pip install this
If you just want to use agami, install the Claude Code plugin — you don't touch this package directly:
/plugin marketplace add AgamiAI/agami-core
/plugin install agami-core@agami
pip install agami-core is for the other two audiences: importing the library into your own
code, or self-hosting the MCP server. Full product docs and the plugin walkthrough live in the
repository.
Install
pip install agami-core # executor + stdio harness (pure-stdlib)
pip install 'agami-core[model]' # + the semantic model (pydantic / sqlglot / pyyaml)
pip install 'agami-core[server]' # + the networked HTTP MCP server (see below)
From a checkout, swap in the editable path — pip install -e 'packages/agami-core[model]'.
Importing the library
The top-level names — semantic_model, mcp_harness, execute_sql, agami_paths — are a
deliberate flat invariant (no parent package, no sys.path juggling), so a consumer's imports
resolve unchanged:
from mcp_harness import TOOLS
import semantic_model
import execute_sql
Module entry points:
python -m mcp_harness # the stdio MCP server (Claude Desktop)
python -m execute_sql --sql-file query.sql # the local query executor
python -m semantic_model.cli # the semantic-model CLI (driven by the `sm` launcher)
python -m mcp_http # the networked HTTP MCP server (see below)
HTTP MCP server ([server]) — early access, in testing
The [server] extra adds a networked MCP transport: the same TOOLS surface as the stdio
server, but over HTTP with OAuth and a small admin console. It's the self-host shape of the hosted
product — deploy it to your own host and a whole team connects their own Claude to one URL,
zero-egress by default (enabling OIDC/SSO adds one outbound call to your identity provider).
🧪 Early access. This team/server layer is usable today but newer than the local single-player path — expect the occasional rough edge, and please report anything broken via a GitHub issue. The local library and stdio server are the stable path.
A minimal local launch (SQLite store, password admin):
PUBLIC_BASE_URL=https://your-host \
AGAMI_SIGNING_SECRET=$(openssl rand -hex 32) \
AGAMI_DB_URL=sqlite:///$PWD/agami.db \
AGAMI_ADMIN_USERNAME=you@example.com AGAMI_ADMIN_PASSWORD=choose-a-strong-one \
python -m mcp_http
The full detail — the auth/access model, OIDC onboarding, the admin console (Activity + Model views), and every environment variable — lives in the docs rather than here:
- Deploy the Docker bundle → deploy/README.md
- Manual install + full env-var reference → docs/self-hosting.md
- Use agami from Claude Desktop (stdio) → docs/mcp-server.md
Links
- Repository & full docs — github.com/AgamiAI/agami-core
- Homepage — agami.ai
- License — fair-code (source-available), the Agami Functional Use License. Self-hosting for your own organization is free; serving people outside it needs a commercial license.
1 maintainer
Release files for agami-core 0.8.7
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | |
|---|---|---|---|
| agami_core-0.8.7.tar.gz | 676.8 kB | Details |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Reset |
|---|---|---|---|---|
| agami_core-0.8.7-py3-none-any.whl | Python 3 | none | any | Details |
Total release size: 1.4 MB
Release files / agami_core-0.8.7.tar.gz
| Download URL | agami_core-0.8.7.tar.gz |
|---|---|
| Size | 676.8 kB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
dd0bc7f4faef8ce990c24dc97f5f3682d10d2708269f99e22f4b829fbdadc37e
|
|
BLAKE2b-256 checksum How to use checksums |
3ebb6aa0ef80be1d34914528d5a08d8af7cb3941e094f18bfd98ba5744e653cc
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
Yes |
| Uploaded via |
twine/6.1.0 CPython/3.13.13
|
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
PyPI Publish Attestation
PyPI verified that this artifact, at this checksum, originated from the publisher listed below.
Signed by GitHub Actions, verified by PyPI on Sep 15, 2026.
Transparency logRelease files / agami_core-0.8.7-py3-none-any.whl
| Download URL | agami_core-0.8.7-py3-none-any.whl |
|---|---|
| Size | 710.7 kB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
550fb87f7c5fa3975c0c3797628a363005b5a315c05cc83f40e5680c6c7738c3
|
|
BLAKE2b-256 checksum How to use checksums |
b93ea1348769bea3669cf11c215b026463cb304afe28e2954ab872d6b851cca5
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
Yes |
| Uploaded via |
twine/6.1.0 CPython/3.13.13
|
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
PyPI Publish Attestation
PyPI verified that this artifact, at this checksum, originated from the publisher listed below.
Signed by GitHub Actions, verified by PyPI on Sep 15, 2026.
Transparency log
