ContextForge is an open-source gateway and registry that brings MCP, A2A, REST, and gRPC tools, agents, and APIs together behind one endpoint. It helps teams discover, govern, observe, authenticate, and route calls across AI infrastructure. The catalogue add-ons support operating and extending this gateway.
Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add IBM/mcp-context-forge --skill observability-audit-sessionsgit clone --depth 1 https://github.com/IBM/mcp-context-forgeWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/ibm/mcp-context-forge/observability-audit-sessions)<a href="https://agentmods.dev/skills/ibm/mcp-context-forge/observability-audit-sessions"><img src="https://agentmods.dev/badge/skills/ibm/mcp-context-forge/observability-audit-sessions/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/ibm/mcp-context-forge/observability-audit-sessions"><img src="https://agentmods.dev/badge/skills/ibm/mcp-context-forge/observability-audit-sessions.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00064 | $0.00904 |
| Opus 5.5 | $0.00026 | $0.00362 |
| Sonnet 5.5 | $0.00013 | $0.00181 |
| Haiku 4.5 | $0.00006 | $0.00090 |
Grade A, and why
observability-audit-sessions scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured today.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 56 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Observability and Audit Trail Session Behavior
Observability writes and audit writes use sessions separate from the main request transaction. This is an intentional trade-off.
Observability Transaction Behavior
Issue #3883 - Separate Session Pattern
The HTTP middleware reuses one observability session per traced request. This means:
- Observability data persists even when the main request fails.
- Traces can show "in progress" or partial states for failed requests.
- Observability writes are NOT atomic with the main request transaction.
- You get visibility into partial failures at the cost of atomicity.
Implementation Details
Write methods (create independent sessions unless callers supply a supported obs_db parameter):
start_trace(),end_trace()start_span(),end_span()add_event(),record_token_usage(),record_metric(),delete_old_traces()
Query methods (use request-scoped sessions):
get_trace(),get_traces(),get_spans(), etc.- These accept a
db: Sessionparameter for RBAC/token scoping.
Context managers (create one independent session for the lifecycle):
trace_span(),trace_tool_invocation(),trace_a2a_request()
Middleware lifecycle: Commit parent traces and HTTP spans before request processing so independent child writers can see them. Commit span and trace completion independently. Roll back caught span failures before attempting trace completion. Keep this session separate from application transactions.
Pattern: SQL instrumentation also reuses a session for each query span in instrumentation/sqlalchemy.py.
Audit Trail Transaction Behavior
Issue #2871 - Separate Session Pattern
AuditTrailService.log_action() (mcpgateway/services/audit_trail_service.py) opens its own SessionLocal() when no db is supplied. It closes and rolls back that session itself. Callers in tool_service.py, resource_service.py, gateway_service.py, prompt_service.py, server_service.py, and admin.py must never pass db=db (the caller's request-scoped session) to log_action(). In admin.py, plugin-view audit logging goes through log_audit(). It is a thin wrapper over log_action() and inherits the same optional-session behavior.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- today First seen · 56 lines · 64 tokens per session scan A 9d52aae7e93b
observability-audit-sessions is a skill published in the GitHub repository IBM/mcp-context-forge (4,832 stars, last pushed today), licensed Apache-2.0. It adds 64 tokens to every session and 904 once invoked, about $0.0003 per session on Opus 5.5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-11-17.
observability-audit-sessions compared with skills from other repositories
| Mod | Type | Stars | Last push | Licence | Scan | Tokens | Works with |
|---|---|---|---|---|---|---|---|
| observability-audit-sessionsIBM/mcp-context-forge | Skill | 4.8k | today | Apache-2.0 | A | 64 | Claude Code, Codex |
| adk-debuggoogle/adk-python | Skill | 22k | today | Apache-2.0 | A | 191 | Claude Code, Codex |
| resolve-handoffomnigent-ai/omnigent | Skill | 11k | today | Apache-2.0 | A | 23 | Claude Code, Codex |
| typescript-testingComposioHQ/composio | Skill | 31k | today | MIT | A | 92 | Claude Code, Codex |
| cli-e2eComposioHQ/composio | Skill | 31k | today | MIT | A | 62 | Claude Code, Codex |
| rubber-duckkirodotdev/KiroCrew | Skill | 5.0k | today | Apache-2.0 | A | 161 | Claude Code, Codex |
| ios-simulatorcallstack/agent-device | Skill | 5.4k | today | MIT | A | 69 | Claude Code, Codex |