One local agent product with two depths.
Meridian Core is the daily cockpit. Meridian Team is the governed depth for approvals, budget, court, and audit.
Website · Install · Contributing
curl -fsSL https://raw.githubusercontent.com/mapleleaflatte03/meridian/main/scripts/install-full.sh | bashcd ~/meridian
./scripts/onboard.shChoose your mode in onboarding:
- Core — daily local work
- Team — governed execution depth
The selected mode is persisted in runtime/onboard_state.json.
After onboarding, supported user-local provider and runtime overrides live in:
~/.meridian/.env~/.meridian/.env.gateway~/.meridian/team.jsonfor role/purpose/team-semantics overrides
Meridian now defaults Team mode to a 7-agent software-delivery team preset. See docs/DEV_TEAM_PRESET.md and docs/TEAM_RUNTIME_CONFIG.md for the source-of-truth config model, supported role swapping, precedence rules, and generated runtime files.
./scripts/core.sh ask "summarize this week"
./scripts/core.sh ask --file src/main.py "review this code"
./scripts/core.sh ask -f a.py -f b.py "compare these files"
./scripts/core.sh session new workbench
./scripts/core.sh ask "continue the plan from earlier"
./scripts/core.sh session show
./scripts/core.sh session export workbench /tmp/core-session
./scripts/core.sh response meta
./scripts/core.sh response page
./scripts/core.sh response export /tmp/core-artifact
./scripts/core.sh chat
./scripts/core.sh context add AGENTS.md docs/plan.md
./scripts/core.sh context list
./scripts/core.sh ask --no-context "one-off task without default project context"
./scripts/core.sh playbook scaffold morning-brief
./scripts/core.sh playbook add release-qa docs/release_qa.md
./scripts/core.sh playbook capture latest-good-output
./scripts/core.sh playbook run release-qa "Focus on regressions only"
./scripts/core.sh playbook every release-qa 3600
./scripts/core.sh playbook daily release-qa 08:30 UTC
./scripts/core.sh playbook schedules
./scripts/core.sh playbook run-scheduled playbook-release-qa
./scripts/core.sh playbook unschedule release-qa
./scripts/core.sh session search "provider-probe-ok"
./scripts/core.sh session resume web_api:exportproof 281
./scripts/core.sh session resume web_api:exportproof 281 --queue
./scripts/core.sh session resume web_api:exportproof 281 --context
./scripts/core.sh session reuse "core-proof-ok" --queue
./scripts/core.sh session reuse "core-proof-ok" --context
./scripts/core.sh doctor
./scripts/core.sh doctor summary
./scripts/core.sh ask --model gpt-4o "quick question"
./scripts/core.sh provider status
./scripts/core.sh provider list
./scripts/core.sh provider use my_profile --model gpt-4o
./scripts/core.sh config show
./scripts/core.sh config set MERIDIAN_BRAIN_MANAGER_MODEL gpt-4o
./scripts/core.sh config get MERIDIAN_BRAIN_MANAGER_MODEL
./scripts/core.sh session archive --older-than 7
./scripts/core.sh schedule status
./scripts/core.sh schedule list
./scripts/core.sh schedule daily morning-brief 08:30 UTC
./scripts/core.sh schedule run morning_brief
./scripts/core.sh channel diagnostics
./scripts/core.sh channel diagnostics telegram
./scripts/core.sh channel diagnostics zalo 10
./scripts/core.sh channel connect list
./scripts/core.sh channel connect scaffold telegram-admin telegram meridian.runtime.v1
./scripts/core.sh channel connect scorecard
./scripts/core.sh web urls
./scripts/core.sh web status
./scripts/core.sh web browse-policy
./scripts/core.sh shell list
./scripts/core.sh shell run repo-status
./scripts/core.sh proof local
./scripts/core.sh proof summary
./scripts/core.sh browse https://example.com
./scripts/core.sh research "git status"
./scripts/core.sh remember my_note "something useful"
./scripts/core.sh recall my_note
./scripts/core.sh agent inspect
./scripts/core.sh channel health
./scripts/core.sh job list
./scripts/core.sh memory overview
./scripts/core.sh inspectTeam mode keeps Core behavior and adds governed execution routes in the local dashboard. These routes are Basic-auth-gated and require onboarding with --mode team.
Runnable example: examples/team-governed-execution.sh
Meridian is split into three layers:
loom/— local agent runtimekernel/— governance engineintelligence/— website, dashboard, proofs, workflows, operator surfaces
- Core first — daily local work without governance overload
- Team when needed — approvals, treasury, court, and audit remain available as deeper control
- Proof visible — inspect runtime and policy posture through live routes
- Local-first — execution and state stay on your machine
./scripts/dev-up.sh
./scripts/dev-down.sh
cargo test --manifest-path loom/Cargo.toml --workspace
cd kernel && python3 -m unittest discover -s kernel/tests -p 'test_*.py'
cd intelligence && python3 -m unittest -v test_gateway_brain_router.py- Why Meridian
- Proofs
- Benchmark lane
- Migration guide
- Core runtime capability matrix
- Onboarding contract
- Team runtime config
These are developer/maintenance commands. First-time users: start with ./scripts/onboard.sh above, not these.
# Start/stop local workspace + gateway (run after onboarding)
./scripts/dev-up.sh
./scripts/dev-down.sh
# Supervisor (auto-restart 18901/19001/8266)
./scripts/dev-supervisor.sh status
# Install only (without auto-start stack)
MERIDIAN_AUTO_START_STACK=0 ./scripts/bootstrap_full.sh
# Loom tests
cargo test --manifest-path loom/Cargo.toml --workspace
# Kernel tests
cd kernel && python3 -m unittest discover -s kernel/tests -p 'test_*.py'
# Intelligence tests
cd intelligence && python3 -m unittest -v test_gateway_brain_router.py- No paywall gate for core runtime/governance usage
- No mandatory commercial checkout path in onboarding
- No closed-source governance module hidden from community review
- Monorepo root: MIT (
LICENSE) kernel/: Apache-2.0 (kernel/LICENSE)loom/andintelligence/: MIT
Canonical source is this monorepo. Legacy repos (meridian-loom, meridian-kernel, meridian-intelligence) are archived mirrors. See docs/REPO_MIGRATION_MAP.md.
Optional support: GitHub Sponsors · Sustainability policy
Meridian positioning contract: docs/MESSAGE_CONTRACT.md